GOOGLE CLOUD GENERATIVE AI

VIRTUAL INTERNSHIP
An internship report submitted in partial fulfillment of the requirements for the award
of the degree of

BACHELOR OF TECHNOLOGY IN
MECANICAL ENGINEERING

Submitted by
KARANAM KALI KRISHNA
(22A31A0320)

Under the guidance of

Mr.Ch.V.Dileep M.Tech.,
Assistant Professor of ME

PRAGATI ENGINEERING COLLEGE

(AUTONOMOUS)
(Approved by AICTE, Permanently Affiliated to JNTUK, KAKINADA & Accredited by NAAC)

1-378, A.D.B. Road, Surampalem, Near Peddapuram-533437

2024-2025
 PRAGATI ENGINEERING COLLEGE
(AUTONOMOUS)

(Approved by AICTE, Permanently Affiliated to JNTUK, KAKINADA & Accredited by NAAC)

1-378, A.D.B. Road, Surampalem, Near Peddapuram-533437

2024-2025
DEPARTMENT OF MECHANICAL ENGINEERING

CERTIFICATE

This is to certify that the Internship report entitled “GOOGLE CLOUD

GENERATIVE AI”, that is being submitted by KARANAM KALI KRISHNA of
III-I bearing (22A31A0320), in partial fulfillment for the award of the Degree of
Bachelor of Technology in MECHANICAL ENGINEERING of Pragati
Engineering College, for the record of bonafide work carried out by him.

Supervisor Head of the Department

Mr.Ch.V.Dileep M.Tech., Dr.G.Avinash M.Tech., Ph.D
Assistant Professor Associate Professor & HOD
Department of ME Department of ME

Bachelor
Computer
M. V. of Technology
Science
Balla
Rajesh, and Engineering
M.Tech.,
Aarathisree (Ph.D)
22A31A4402
Assoc. Prof & HoD – CSE(DS)
COMPLETION CERTIFICATE
 Student’s Declaration
I, a student of
Program, Reg. No. of the Department of
College do hereby declare that I have completed the mandatory internship from
to in (Name of
the intern organization) under the Faculty Guideship of
(Name of the Faculty Guide), Department of
,
(Name of the College)

(Signature and Date)

 Official Certification
This is to certify that (Name of
the student) Reg. No. has completed his/her Internship in
(Name of the Intern Organization) on
(Title of the Internship) under my
supervision as a part of partial fulfillment of the requirement for the
Degree of in the Department of
(Name of the College).

This is accepted for evaluation.

(Signatory with Date and Seal)

Endorsements

Faculty Guide

Head of the Department

Principal
 ACKNOWLEDGEMENT

It gives me immense pleasure to express a deep sense of gratitude to my Guide, Department of

Mr.Ch.V.Dileep, Mechanical Engineering), for his wholehearted and invaluable guidance
throughout this report. Without his sustained and sincere efforts, this report would not have taken
this shape. He encouraged and helped me overcome the various challenges encountered at different
stages of the internship.

I would also like to sincerely thanks, Dr. G. Avinash, Head of the Department of Mechanical
Engineering, for providing all the necessary facilities that contributed to the successful completion
of this report.

I take this opportunity to thank our beloved Director (Academics) Dr. K. Satyanarayana and
Principal Dr. G. Naresh for their continuous support throughout the internship project and for
giving me the opportunity to undertake this valuable experience.

My heartfelt thanks to Dr. P. Krishna Rao, Chairman of Pragati Engineering College, for
providing an inspiring environment to work and learn.

I extend my gratitude to the entire Department of Mechanical Engineering for their guidance
and support during the course of this internship.

Finally, I would like to thank my friends and family members for their unwavering help and
encouragement throughout this project. Their support has been invaluable in completing this report
successfully.

KARANAM KALI KRISHNA

(22A31A0320)
 Contents

1. Introduction to Generative AI
1.1 Introduction 1
1.2 Purpose/Importance 1- 2
1.3 Modules Overview 2- 6
2. Fundamentals of Generative AI
2.1 Large Language Models 7
2.2 Responsible AI and Ethical AI Principles 7-8
2.3 Model Training and Prompt Design 8
3. Advanced Concepts
3.1 Transformer Models 9 - 10
3.2 Gemini AI and Vertex AI Studio 10
3.3 Image and Text Generation 11
4. Internship Activities
4.1 Course Completion Log 12
4.2 Projects and Assignments Completed 13 - 32
5. Outcomes and Skills Acquired
5.1 Technical Skills 33 - 34
5.2 Communication and Managerial Skills 34
5.3 Overall Internship Evaluation 34 - 35
6. Conclusion 36
 CHAPTER 1: INTRODUCTION TO GENERATIVE AI

1.1 Introduction
Generative AI is a transformative subset of artificial intelligence focused on creating new, original
content—whether it's text, images, audio, or code—by analyzing and learning patterns from vast
datasets. Unlike traditional AI models that simply classify or predict based on input data, generative
AI has the ability to generate entirely new, creative outputs. This technology forms the backbone of
applications such as chatbots, image generators, and code autocompletion tools.
During my internship, I was fortunate to gain hands-on experience with Google Cloud’s Vertex
AI and Gemini AI, exploring a wide array of modules that focused on large language models
(LLMs) and their real-world applications. The tools and models I worked with have been pivotal
in enhancing my understanding of how AI models can be deployed to solve complex problems,
automate processes, and drive innovation across industries.
Suggested Image: A diagram explaining Generative AI, showcasing its subsets like text
generation, image generation, and code generation. This can include a simple flowchart depicting
how large datasets are processed by models like Transformer to generate outputs.

1.2 Purpose/Importance
The primary goal of this internship was to provide participants with both foundational and
advanced knowledge in Generative AI. It delved deeply into the ethical implications of AI while
simultaneously offering practical insights into model development and deployment processes.
The integration of Vertex AI with Gemini AI allowed me to experience firsthand the challenges
and opportunities associated with training and deploying AI models in real-world applications.
The curriculum consisted of three core courses:
 Introduction to Generative AI
This learning path provides an overview of generative AI concepts,
from the fundamentals of large language models to responsible
AI principles.
 Intermediate: Gemini for Google Cloud Learning
The Gemini for Google Cloud learning path provides examples of how
Gemini can help make engineers of all types more efficient in their daily activities.
Gemini provides a natural language chat interface which you can
quickly chat with to get answers to cloud questions, or receive guidance on best practices.

Page No: 1
  Advanced: Generative AI for Developers
A Generative AI Learning Path with a technical focus, built for
App Developers, Machine Learning Engineers, and Data Scientists.
Each course was designed to progressively build my knowledge, starting from
basic concepts and culminating in advanced topics such as transformer architectures,
multimodal AI, and responsible AI design principles.

1.3 Modules Overview

Throughout the internship, I completed several modules, each designed to cover key aspects of
Generative AI:
 Prompt Design: Understanding how to create prompts that guide AI models to generate
relevant outputs.
 Responsible AI Principles: Ensuring AI models are developed and deployed in ways
that promote fairness and avoid bias.
 Transformer and BERT Models: Gaining insights into advanced architectures like
Transformers, which power models like GPT and BERT.
These modules provided the theoretical and practical foundation to not only understand the
underlying mechanics of AI but also apply this knowledge in real-world contexts.

Beginner: Introduction to Generative AI Learning Path

Page No: 2
Intermediate: Gemini for Google Cloud Learning Path

Page No: 3
Page No: 4
Advanced: Generative AI for Developers Learning Path

Page No: 5
Page No: 6
 CHAPTER 2: FUNDAMENTALS OF GENERATIVE AI

2.1 Large Language Models (LLMs)

LLMs are at the heart of Generative AI, and the internship provided in-depth exposure to their
architecture. LLMs are trained on massive datasets and can perform tasks such as text generation,
language translation, and even summarization by learning patterns in the data. Key topics I explored
included:
 Pre-training and Fine-tuning: Pre-training a model on a general dataset, followed by fine-
tuning it on a specific task or domain.
 Text Generation: Using LLMs to generate coherent and contextually accurate text based on
input prompts.
Understanding LLMs not only allowed me to appreciate their current capabilities but also to
anticipate their evolving applications in various domains.

2.2 Responsible AI and Ethical AI Principles

Ethical AI development is crucial in today's world, and this internship underscored the importance

Page No: 7
of ensuring fairness, transparency, and accountability in AI solutions. Through Google’s AI
Principles, I learned how to design AI systems that mitigate bias, uphold privacy, and ensure
equitable outcomes for all users.
This module emphasized:
 Avoiding AI Bias: Techniques to reduce unintended biases in AI models by ensuring diverse
training data and rigorous testing.
 Ethical AI Deployment: Strategies to ensure that AI tools are deployed responsibly,
particularly in sensitive applications like healthcare and finance.

2.3 Model Training and Prompt Design

One of the more practical aspects of the internship involved Prompt Design—the process of crafting
precise prompts to direct AI models like Gemini AI in generating accurate responses. Understanding
how to structure prompts allowed me to maximize the effectiveness of the models and generate
more meaningful outputs.
Key learnings included:
 Effective Prompt Structuring: How to guide AI to generate desired responses.
 Iterative Testing: Refining prompts based on model feedback to optimize results.

Page No: 8
 CHAPTER 3: ADVANCED CONCEPTS

3.1 Transformer Models

The Transformer architecture marks a groundbreaking advancement in natural language

processing (NLP) and generative models, fundamentally reshaping the way AI understands and
generates text. Unlike previous models like RNNs (Recurrent Neural Networks) and LSTMs (Long
Short-Term Memory), which processed input sequentially, the Transformer model processes input in
parallel, dramatically increasing computational efficiency and accuracy. During the internship, I had
the opportunity to delve deeply into the intricacies of Transformer models, with a specific focus on
BERT (Bidirectional Encoder Representations from Transformers).
Transformers use a unique self-attention mechanism that enables them to weigh the importance
of different words in a sentence. This feature allows the model to focus on different parts of the
input depending on context, making it especially powerful for understanding and generating
coherent, contextually relevant text. Here are the key concepts I explored:
 Self-Attention Mechanism: A core aspect of the Transformer model, this mechanism
allows the model to evaluate the relationship between each word in a sentence, paying more
attention to certain words depending on the context. This is vital for tasks like translation,
summarization, and text generation, where understanding the relationship between words
and phrases is critical.
For example, in a sentence like "The dog chased the cat," the model can understand that "the cat"
is the object being chased, even though the word "chased" appears earlier in the sentence. This
attention to context dramatically improves the quality of text generation and comprehension tasks.
 BERT’s Impact on NLP: BERT, an application of the Transformer architecture,
represents a significant leap forward in NLP because of its ability to process text
bidirectionally, analyzing both the left and right contexts of a word simultaneously. This
bidirectional analysis allows BERT to have a deeper understanding of ambiguous words
and phrases, improving its performance on tasks like question answering, text
classification, and language inference. BERT's pre-training on large datasets followed by
fine-tuning on specific tasks has set new benchmarks for NLP models, proving to be a game
changer in AI applications.

Page No: 9
The experience of working with BERT models not only solidified my understanding of advanced
NLP techniques but also provided me with the skills to apply these models in real-world scenarios
like language translation, content generation, and chatbot development.

3.2 Gemini AI and Vertex AI Studio

One of the most exciting and enriching aspects of the internship was working hands-on with Gemini
AI within the Vertex AI Studio environment. Gemini AI is a state-of-the-art multimodal AI system
designed to handle both text and image inputs, making it an extremely versatile tool for developing
a wide range of applications. This dual capability enables it to perform tasks such as image
captioning, text generation, image recognition, and multimodal translation, bridging the gap
between different data modalities in innovative ways.
Gemini AI stands out for its ability to integrate text and image data seamlessly, which is
particularly valuable for industries like:
 Media and Entertainment, where it can generate content (both text and visuals) based on
user preferences.
 Marketing, where it helps create personalized content for customers.
 Healthcare, where it assists in analyzing medical images alongside patient records for a
more comprehensive diagnosis.
In Vertex AI Studio, I learned how to:
 Build and deploy AI models: Using pre-built pipelines, I developed models for both text
and image processing. These models were trained on large datasets, fine-tuned for specific
tasks, and deployed in real-world scenarios.
 Manage and scale AI applications in real-time: I explored how to deploy these models
in a scalable cloud environment, allowing them to handle large amounts of data efficiently.
Real-time monitoring features provided insights into model performance, resource
allocation, and operational costs, ensuring that applications were running optimally.
This hands-on experience in Vertex AI Studio was invaluable in understanding the end-to-end
process of AI model development—from training and tuning to deployment and scaling. It
provided a real-world perspective on the challenges and solutions associated with managing AI in
a cloud-based environment, especially for large-scale applications requiring high computational
resources.

Page No: 10
3.3 Image and Text Generation

The ability to generate both text and images from AI models is one of the most fascinating and
revolutionary aspects of Generative AI. This part of the internship allowed me to explore multimodal
models that can understand and synthesize text and image data in meaningful ways.
 Text-to-Image Generation: Using pre-trained models, I learned how to generate images
from textual descriptions. This technology has far-reaching applications, from creating
visual content for marketing campaigns to generating realistic designs for virtual
environments. By providing a detailed textual prompt, the model was able to produce
images that closely matched the description. For example, when asked to generate an image
of "a sunset over a mountain range with purple skies," the model produced a stunning visual
that aligned with the input prompt.
 Image Captioning: One of the more exciting applications of multimodal AI was learning
how to create captions for images. The model analyzed images and generated contextually
accurate descriptions, demonstrating AI’s capacity to understand and explain visual
content. This is particularly useful in areas such as accessibility (where AI can generate
descriptions for visually impaired users), content moderation, and automated content
generation for social media platforms.
For instance, when shown an image of a cat sitting by a window, the AI-generated caption was “A
cat sitting on a windowsill looking out at the sky.” This not only showed the model’s ability to
identify objects but also its understanding of context and relationships between objects.
Through these tasks, I experienced firsthand how multimodal AI is bridging the gap between
different forms of data—opening up new creative and practical applications. From generating
artwork and writing stories to developing AI-powered assistants that interact with the world
through both words and visuals, the potential for innovation is boundless.

Page No: 11
 CHAPTER 4: INTERNSHIP ACTIVITIES

4.1 Course Completion Log

 Completed Introduction to Generative AI, earning 5 badges.
 Completed Gemini for Google Cloud Learning, earning 8 badges.
 Completed Advanced: Generative AI for Developers, earning 11 badges.

4.2 Projects and Assignments Completed

 Text Generation Model: Built a model using Gemini AI to generate coherent and
contextually accurate text.
 Image Captioning System: Developed a system that generates descriptive captions for
images, combining text and image inputs.
 Real-Time AI Solutions: Implemented AI solutions using Vertex AI Studio for tasks like
real-time data analysis and image recognition.

Page No: 12
 ACTIVITY LOG FOR THE FIRST WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Created profile on Familiarized with platform

Day – 1
Google Cloud Skills Boost features.

Completed 'A Tour of Google Cloud Learned basic cloud

Hands-on Labs' ~45min. services.
Day - 2

Completed 'Introduction to Understood Generative AI

Day – 3
Generative AI' ~45min fundamentals.

Completed 'Introduction to Large Learned about LLMs and

Day – 4
Language Models' ~30min their applications.

Completed 'Introduction to Grasped importance of

Day – 5 responsible AI.
Responsible AI' ~30min

Started 'Prompt Design in Vertex AI' Learned prompt

Day –6 engineering basics.
~1 hr 30min

Page No: 13
 WEEKLY REPORT
WEEK – 1 (From Dt to Dt )

Objective of the Activity Done:

Get familiar with Google Cloud services and explore generative AI basics. Complete
foundational hands-on labs to gain a practical understanding of cloud infrastructure.

Detailed Report:

In the first week, I focused on getting familiar with the Google Cloud environment by creating a
profile and navigating through the platform. The hands-on tour provided a clear overview of
Google Cloud services, which would serve as a foundation for more complex tasks in the following
weeks.

I also completed a set of introductory modules on Generative AI, where I learned about the
underlying mechanisms and applications of these AI systems, including Large Language Models
(LLMs). LLMs were particularly fascinating due to their ability to process vast amounts of textual
data and generate coherent outputs. The Responsible AI module offered insights into ethical
considerations, ensuring AI systems are deployed in a fair and unbiased manner.

One of the major learnings this week was the critical importance of prompt engineering, which I
began exploring in Vertex AI. Crafting effective prompts for AI models is essential in guiding their
behaviour and obtaining meaningful results. The introduction to prompt design provided a good
mix of theory and practice, giving me a solid base for more advanced work in subsequent weeks.

Page No: 14
 ACTIVITY LOG FOR THE SECOND WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Continued 'Prompt Design in Advanced prompt

Day – 1
Vertex AI' ~2 hrs. design techniques.

Completed 'Responsible AI: Applied AI principles

Day - 2 Applying AI Principles' ~2 hrs. to real-world scenarios.

Completed 'Gemini for Learned how Gemini

Day – 3 Application Developers' ~1 hr 30min. aids app development.

Completed 'Gemini for Cloud Explored Gemini's impact

Day – 4
Architects' ~1 hr. on infrastructure.

Started 'Gemini for Data Scientists Understood Gemini’s role

Day – 5 and Analysts' ~2 hrs. in data science.

Continued 'Gemini for Hands-on data analysis

Day –6 Data Scientists and Analysts' ~1 hr. with Gemini.

Page No: 15
 WEEKLY REPORT
WEEK – 2 (From Dt to Dt )
Objective of the Activity Done:

Understand the architecture and workings of LLMs like BERT and GPT. Explore how to train,
fine-tune, and deploy LLMs for various NLP tasks.

Detailed Report:

Week 2 was focused heavily on expanding my knowledge in prompt design with Vertex AI. I
continued building upon the basic concepts I had learned in Week 1 and delved into more advanced
prompt engineering techniques. These techniques will allow me to fine-tune interactions with AI
models and ensure more precise results, which is critical in any application involving AI-driven
text or content generation.

Additionally, I began exploring Google’s Gemini model, a comprehensive AI framework designed

to assist in multiple domains such as application development, cloud architecture, and data science.
I was particularly impressed by the versatility of Gemini, which can be applied across various
stages of the application development pipeline. The courses demonstrated how developers and
architects can utilize Gemini to accelerate the creation of scalable, high-performance applications
while maintaining efficient cloud operations. By the end of the week, I had also started working
on the application of Gemini for data scientists, particularly in data analysis and model-building
processes. This laid the groundwork for more practical exercises involving real datasets.

Understanding the principles of responsible AI became even more critical this week, especially
when I explored the "Responsible AI: Applying AI Principles" module. This helped me see how
ethical AI development can be integrated into real-world scenarios to mitigate bias and enhance
transparency.

Page No: 16
 ACTIVITY LOG FOR THE THIRD WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Finished 'Gemini for Data Scientists Applied Gemini to

Day – 1
and Analysts' ~1 hr. customer data analysis.

Completed 'Gemini for Network Learned about network

Day - 2
Engineers' ~1 hr. management with Gemini.

Completed 'Gemini for Security Enhanced cloud security

Day – 3
Engineers' ~45min. using Gemini.

Completed 'Gemini for DevOps Streamlined DevOps

Day – 4
Engineers' ~1 hr 15min. workflows with Gemini.

Started 'Gemini for Explored Gemini for

Day – 5
End-to-End SDLC' ~1 hr 30min. software lifecycle.

Completed 'Gemini for Applied SDLC practices

Day –6
End-to-End SDLC' ~1 hr. with Gemini.

Page No: 17
 WEEKLY REPORT
WEEK – 3 (From Dt to Dt )
Objective of the Activity Done:

Learn about responsible AI development practices, focusing on fairness, transparency, and

ethical considerations. Implement AI models adhering to these principles using Google Cloud
tools.

Detailed Report:

In Week 3, the emphasis shifted towards how Google’s Gemini model could be leveraged by
different engineering roles, such as network engineers, security professionals, and DevOps
engineers. I found it fascinating how Gemini could help automate and simplify complex workflows
for these professionals. For instance, Gemini’s capabilities for network engineers demonstrated
how AI can monitor and optimize network performance in real-time, reducing human intervention
and potential errors. Similarly, the modules on security engineering highlighted how Gemini
enhances cloud security, particularly in identifying and mitigating potential threats in cloud
infrastructure.

The DevOps module introduced me to ways in which Gemini can streamline the deployment of
applications through continuous integration and continuous deployment (CI/CD) pipelines. This is
especially useful for ensuring faster and more reliable software releases. I also completed the "End-
to-End Software Development Lifecycle (SDLC)" module, which provided an in-depth
understanding of how AI can assist in managing every phase of the SDLC, from planning and
development to testing and deployment. I applied these concepts in practical exercises, where I
built and deployed small-scale applications using Gemini’s advanced tools.

By the end of this week, I had gained a solid grasp of how AI can support various aspects of cloud
and application infrastructure, as well as the importance of integrating AI into DevOps workflows
for greater efficiency and automation.

Page No: 18
 ACTIVITY LOG FOR THE FORTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Completed 'Develop GenAI Apps Built apps using Gemini

Day – 1
with Gemini & Streamlit' ~2 hrs. and Streamlit.

Continued 'Develop GenAI Apps Explored Streamlit app

Day - 2
with Gemini & Streamlit' ~2 hrs. deployment.

Finished 'Develop GenAI Apps Deployed Streamlit apps

Day – 3
with Gemini & Streamlit' ~1 hr 45min. in Cloud Run.

Completed 'Introduction to Image Explored diffusion models

Day – 4
Generation' ~30min. for image creation.

Completed 'Attention Mechanism' Learned about attention in

Day – 5
~30min. neural networks.

Completed 'Encoder-Decoder Implemented encoder-

Day –6
Architecture' ~30min. decoder model.

Page No: 19
 WEEKLY REPORT
WEEK – 4 (From Dt to Dt )

Objective of the Activity Done:

Master prompt engineering techniques to optimize generative AI outputs. Experiment with

different prompting methods and assess their impact on model responses.

Detailed Report:

Week 4 was particularly hands-on, focusing on building generative AI applications using Gemini
and Streamlit. These projects were exciting because they allowed me to apply AI concepts in real-
world applications. Streamlit, combined with Gemini’s powerful APIs, enabled me to quickly
prototype and deploy machine learning applications. Over the course of several days, I learned how
to deploy these apps using Cloud Run, which offered scalability and simplicity in managing AI
workloads.

Additionally, this week introduced me to the fascinating world of image generation using diffusion
models. Diffusion models are a recent advancement in AI-driven image creation, where a model
can generate high-quality images by iteratively refining noisy input data. I gained hands-on
experience with basic image generation tasks and how to utilize these models in creative
applications.

I also explored the attention mechanism and encoder-decoder architecture. The attention
mechanism is key to enhancing the performance of AI models, particularly in natural language
processing (NLP) tasks. Understanding the encoder-decoder architecture, which is fundamental in
many sequence-to-sequence tasks like translation and summarization, further deepened my
knowledge of AI model structures.

Page No: 20
 ACTIVITY LOG FOR THE FIFTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Completed 'Transformer Models Explored BERT and

Day – 1
and BERT Model' ~30min. Transformer models.

Completed 'Create Image Created models for

Day - 2
Captioning Models' ~30min. image captioning.

Completed 'Introduction to Prototyped models in

Day – 3
Vertex AI Studio' ~2 hrs. Vertex AI Studio.

Started 'Vector Search Learned about vector

Day – 4
and Embeddings' ~1 hr. search techniques.

Finished 'Vector Search Implemented text

Day – 5
and Embeddings' ~1 hr. embeddings for search.

Started 'Inspect Rich Documents Applied multimodal

Day –6 retrieval-augmented
with Gemini Multimodality' ~2 hrs.
generation techniques.

Page No: 21
 WEEKLY REPORT
WEEK – 5 (From Dt to Dt )
Objective of the Activity Done:

Gain expertise in using BERT for NLP tasks and building image captioning models. Explore
MLOps to streamline model deployment, monitoring, and management.

Detailed Report:

Week 5 focused on exploring advanced machine learning models, including Transformer and
BERT models. Transformers are the backbone of many modern NLP applications, and learning
how they work at a deep level opened up new ways to implement state-of-the-art text generation,
translation, and summarization models. Specifically, BERT’s pre-training and fine-tuning
capabilities allow it to adapt to a wide range of tasks, which I practiced through hands-on labs.

Additionally, I built image captioning models, which combine image analysis with natural
language generation to create descriptive captions for images. This task highlighted the power of
combining visual and textual data to generate more comprehensive AI solutions.

I also dived deeper into machine learning operations (MLOps) through Vertex AI Studio, which
provided a unified interface for developing, deploying, and managing AI models. This week’s
focus on vector search and embeddings also proved invaluable for improving search engine
capabilities by representing text as numerical vectors, enabling more relevant and contextual
results.

Page No: 22
 ACTIVITY LOG FOR THE SIXTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Completed 'Gemini for Multimodal Explored multimodal

Day – 1
AI Applications' ~1 hr. AI for text and images.

Worked with rich

Finished 'Inspect Rich Documents documents for
Day - 2
with Gemini Multimodality' ~2 hrs. multimodal tasks.

Completed 'Prompt Engineering Applied retrieval-

Day – 3
with Retrieval-Augmented Generation' augmented generation.
~1 hr 30min.

Finished 'Using Advanced Prompting Advanced prompt tuning

Day – 4
Techniques in Vertex AI' ~1 hr. with Vertex AI.

Continued working with Developed search engines

Day – 5
Vector Search applications. with embeddings.

Completed 'Vector Search Finalized vector search

Day –6
with Gemini' ~2 hrs. implementations.

Page No: 23
 WEEKLY REPORT
WEEK – 6 (From Dt to Dt )

Objective of the Activity Done:

Dive into multimodal AI applications combining text and image data. Implement vector search
algorithms to enhance contextual understanding in AI-driven search engines.

Detailed Report:

In Week 6, I took a deep dive into multimodal AI with Gemini, learning how to combine different
data modalities such as text, images, and structured data for more comprehensive applications.
Gemini’s multimodal capabilities were especially powerful in handling complex AI tasks that
required understanding and processing different types of inputs simultaneously.

Multimodal AI techniques are crucial in areas like autonomous driving, where both visual and
sensor data are combined, or in rich document analysis, where text, images, and metadata must be
processed together. I also spent a significant amount of time working with retrieval-augmented
generation, which helps in retrieving relevant content from large databases and combining it with
prompt generation tasks.

During this week, I also completed my work on vector search, applying text embeddings to build
advanced search engines. Vector search enables more contextual and semantically rich searches,
which was an exciting addition to my skill set as it can be used to improve search results in AI-
driven platforms.

Page No: 24
 ACTIVITY LOG FOR THE SEVENTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Completed 'Building Generative AI Applied Generative AI for

Day – 1
Business Applications' ~2 hrs. business use cases.

Completed 'Gemini for Financial Applied AI in the

Day - 2
Services' ~1 hr 15min. financial industry.

Explored 'Vertex AI for Custom Built custom machine

Day – 3
ML Models' ~1 hr. learning models.

Completed 'Vertex AI for Model Implemented monitoring

Day – 4
Monitoring' ~1 hr. tools for models.

Completed 'Vertex AI for Feature Built features with

Day – 5
Engineering' ~1 hr. Vertex AI.

Explored 'Vertex AI Pipelines' Integrated pipelines for

Day –6
~1 hr 30min. automated workflows.

Page No: 25
 WEEKLY REPORT
WEEK – 7 (From Dt to Dt )

Objective of the Activity Done:

Apply generative AI to real-world business problems and integrate it into cloud-based solutions.
Learn to build and monitor custom AI models using Vertex AI.

Detailed Report:

In Week 7, I focused on the application of generative AI for business solutions. The "Building
Generative AI Business Applications" module provided a comprehensive understanding of how
AI can be applied to solve real-world business problems. I learned about deploying AI-powered
solutions for tasks like customer service automation, marketing content generation, and financial
services optimization. The financial services module provided a deeper look into how AI can be
used in areas such as fraud detection, credit scoring, and automated trading.

Additionally, I explored Vertex AI’s capabilities for building and deploying custom machine
learning models. Learning about model monitoring and feature engineering was particularly
interesting, as it helped me understand the importance of keeping AI models operational and
performing well over time. Vertex AI Pipelines provided the ability to automate ML workflows,
ensuring that tasks like data preprocessing, model training, and evaluation could be handled
efficiently.

Page No: 26
 ACTIVITY LOG FOR THE EIGTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Completed 'Generative AI Built AI-driven

Day – 1
for Developers' ~2 hrs. applications for developers.

Continued 'Generative AI Explored developer tools

Day - 2
for Developers' ~2 hrs. for AI integration.

Finished 'Generative AI Deployed developer-

Day – 3
for Developers' ~1 hr. focused AI applications.

Completed 'Google Cloud Learned cloud

Day – 4
Infrastructure Basics' ~1 hr. infrastructure fundamentals.

Explored 'Google Kubernetes Worked with containerized

Day – 5
Engine (GKE)' ~1 hr 30min. applications.

Explored 'Cloud Functions and Built serverless apps

Day –6
Serverless Applications' ~2 hrs. with Cloud Functions.

Page No: 27
 WEEKLY REPORT
WEEK – 8 (From Dt to Dt )

Objective of the Activity Done:

Develop advanced generative AI applications for developers and explore cloud infrastructure
tools like GKE and Cloud Functions. Learn to manage containerized and serverless applications.

Detailed Report:

In Week 8, I focused on more advanced concepts in generative AI and Google Cloud infrastructure.
The "Generative AI for Developers" module was extremely insightful, offering practical tools and
strategies for integrating AI-driven features into developer applications. I was able to build
applications that leveraged AI to provide personalized user experiences and automate routine
development tasks. It was interesting to see how developers could use AI to streamline their
workflow and improve productivity.

I also spent time exploring Google Cloud infrastructure basics and services such as the Google
Kubernetes Engine (GKE) and Cloud Functions. These tools are essential for deploying
containerized and serverless applications, respectively. Working with GKE introduced me to the
concept of managing containerized applications at scale, while Cloud Functions offered a
serverless approach to handling individual functions in response to cloud events.

Page No: 28
 ACTIVITY LOG FOR THE NINTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Continued work on hands-on project Finalized project

Day – 1
using Generative AI tools. components.

Integrated AI models with Deployed application

Day - 2
Cloud-based infrastructure. using Cloud Run.

Tested AI model performance Enhanced model

Day – 3
and debugging. performance.

Completed final hands-on Refined AI

Day – 4
exercises with Vertex AI. model integration.

Reviewed and polished AI Ensured functionality

Day – 5
applications for final submission. and user experience.

Submitted hands-on project Completed course

Day –6
for evaluation. assignments.

Page No: 29
 WEEKLY REPORT
WEEK – 9 (From Dt to Dt )

Objective of the Activity Done:

Apply all learned techniques in a hands-on project, integrating AI models with cloud
infrastructure. Test, debug, and finalize the project for submission.

Detailed Report:

In the penultimate week, the focus was on completing the hands-on project, where I applied the
generative AI techniques learned over the previous weeks. I worked extensively with Google’s
cloud infrastructure to deploy and manage the application, using services like Cloud Run to ensure
scalability and performance. The project involved integrating AI models into a real-world use case,
where I built a fully functional application that solved a specific problem using AI-driven features.

This week was also about testing and debugging the model to ensure it performed well in a live
environment. I spent a considerable amount of time fine-tuning the application to enhance both the
AI model’s accuracy and the overall user experience. The final touches included refining the
design, ensuring all cloud services worked seamlessly, and preparing the project for submission.

Page No: 30
 ACTIVITY LOG FOR THE TENTH WEEK

Day Person In-

Brief description of the daily
& Learning Outcome Charge
activity
Date Signature

Final review of all modules Reviewed key concepts

Day – 1
and learning paths. and techniques.

Completed evaluation of Evaluated the performance

Day - 2
submitted course evaluation. of hands-on project.

Submitted Google Cloud Skills Uploaded profile

Day – 3
Boost Profile to Eduskills Portal. for verification.

Completed final quizzes Gained feedback on

Day – 4
and assessments. learning progress.

Submitted the course Finalized course

Day – 5
completion report to Eduskills. completion.

Celebrated completion of 24 Achieved all course

Day –6
badges and learning paths. milestones.

Page No: 31
 WEEKLY REPORT
WEEK – 10 (From Dt to Dt )

Objective of the Activity Done:

Review key concepts, finalize all deliverables, and complete the course assessments. Submit the
project for evaluation and earn certification for the course.

Detailed Report:

The final week was all about consolidation and reflection. I reviewed the entire course, revisiting
key modules to ensure I had a comprehensive understanding of generative AI and its applications
across various domains. I also evaluated the hands-on project submitted in Week 9 to assess how
well it met the course requirements. The feedback provided on my project helped me understand
areas where I excelled and areas for improvement.

After completing the final assessments and quizzes, I submitted my Google Cloud Skills Boost
profile and course completion report to Eduskills for verification. This marked the successful
completion of the 10-week course, earning all 24 badges across various learning paths. It was a
rewarding experience that not only improved my technical skills but also gave me a deeper
understanding of AI's transformative potential across industries.

Page No: 32
 CHAPTER 5: OUTCOMES AND SKILLS ACQUIRED

The Generative AI Virtual Internship provided an invaluable opportunity to grow both

technically and professionally. By participating in this program, I was able to develop a
comprehensive understanding of cutting-edge AI technologies and enhance my communication
and management abilities. Below is a detailed elaboration of the key outcomes and skills I acquired
during this internship.

5.1 Technical Skills

Throughout the internship, I gained extensive technical knowledge and hands-on experience with
advanced AI concepts, tools, and methodologies. Some of the core technical skills I acquired
include:

 Mastered Transformer Models and BERT: I developed an in-depth understanding of the

Transformer architecture, particularly focusing on BERT (Bidirectional Encoder
Representations from Transformers). This allowed me to grasp how these models utilize
the self-attention mechanism to process input sequences in parallel, enhancing efficiency
in natural language processing (NLP) tasks. By working directly with BERT, I learned how
bidirectional analysis enables the model to understand context more effectively than
traditional models, making it superior for tasks such as text classification, summarization,
and question-answering.
 Proficiency in Vertex AI for Model Training and Deployment: I gained practical
experience using Google Cloud’s Vertex AI platform to build, train, and deploy machine
learning models. Through the internship, I became proficient in managing data pipelines,
setting up training environments, fine-tuning pre-trained models, and scaling AI
applications in a cloud-based infrastructure. This hands-on exposure to Vertex AI has given
me a strong foundation in cloud-based AI solutions, which is critical for deploying AI
models at scale in production environments.
 Prompt Design and Responsible AI Practices: Designing effective prompts is crucial for
steering AI models towards generating the desired outputs. I honed my skills in prompt
engineering by experimenting with different approaches to elicit accurate, contextually
relevant responses from models like Gemini AI. Additionally, the internship emphasized
the importance of responsible AI practices, such as ensuring fairness, reducing bias, and
maintaining transparency. These are key elements when developing and deploying AI

Page No: 33
 systems, particularly in sensitive areas like healthcare, finance, and social media. This
understanding has positioned me to implement AI in a manner that is both innovative and
ethically sound.

5.2 Communication and Managerial Skills

The internship also significantly enhanced my communication and managerial abilities. Effective
communication and project management are critical components in the development and
implementation of AI solutions, especially when working in multidisciplinary teams.

 Improved Ability to Explain Complex AI Concepts: One of the most important non-
technical skills I developed was the ability to explain complex AI concepts in both written
and oral formats. Throughout the internship, I was tasked with summarizing technical
topics such as Transformer architectures, prompt engineering, and multimodal AI in
ways that could be easily understood by both technical and non-technical audiences.
Whether through presentations or written reports, I gained confidence in breaking down
complex topics into digestible insights, making AI technology more accessible and
understandable.
 Enhanced Project Management Skills: The internship involved managing multiple
modules, assignments, and projects simultaneously, often with tight deadlines. I improved
my time management and organizational skills by developing a structured approach to
tackling each task. This involved setting priorities, creating timelines, and consistently
meeting deadlines while maintaining the quality of work. These project management skills
will be critical as I move forward in my career, particularly in environments where
managing multiple AI projects in parallel is common.

5.3 Overall Internship Evaluation

This internship has been a deeply enriching experience, both technically and professionally. It has
significantly enhanced my technical proficiency in Generative AI, particularly in areas such as
Transformer models, BERT, Gemini AI, and Vertex AI Studio. The hands-on experience I
gained with these cutting-edge technologies has deepened my understanding of how AI models are
built, trained, and deployed in cloud environments.

Page No: 34
Furthermore, this experience heightened my awareness of the ethical considerations necessary
when deploying AI technologies. The emphasis on Responsible AI throughout the program helped
me realize the broader implications of AI in society, and the importance of developing systems that
are fair, transparent, and accountable. This awareness has equipped me with the knowledge and
tools to implement AI systems in a responsible and ethical manner, ensuring that they deliver
positive outcomes for users and stakeholders alike.
In conclusion, the internship has not only refined my technical capabilities but also improved my
communication, organizational, and ethical decision-making skills. I feel well-prepared to
contribute to the field of AI and to drive innovation while maintaining a focus on responsible AI
practices.

Page No: 35
 CHAPTER 6: CONCLUSION

The Generative AI Virtual Internship offered me a strong technical foundation, focusing on key areas
such as Large Language Models (LLMs), Transformers, and multimodal AI. Through this program, I
gained hands-on experience with advanced AI concepts, honing my skills in building and fine-tuning AI
models.
However, the internship also placed significant emphasis on the ethical implications of AI. I learned how
to approach AI development responsibly, considering factors like bias, fairness, and the societal impact
of AI systems. This ethical framework has equipped me with the knowledge to not only create innovative
solutions but also ensure that they are aligned with ethical standards and contribute positively to society.
Overall, this internship has provided me with a well-rounded understanding of both the technical and
ethical aspects of AI, preparing me to make meaningful contributions to the field of artificial intelligence.

Page No: 36
 Student Self Evaluation of the Short-Term Internship

Student Name: Registration No:

Term of Internship: From: To :

Date of Evaluation:

Organization Name & Address:

Please rate your performance in the following areas:

Rating Scale: Letter grade of CGPA calculation to be provided

1 Oral communication 1 2 3 4 5
2 Written communication 1 2 3 4 5
3 Proactiveness 1 2 3 4 5
4 Interaction ability with community 1 2 3 4 5
5 Positive Attitude 1 2 3 4 5
6 Self-confidence 1 2 3 4 5
7 Ability to learn 1 2 3 4 5
8 Work Plan and organization 1 2 3 4 5
9 Professionalism 1 2 3 4 5
10 Creativity 1 2 3 4 5
11 Quality of work done 1 2 3 4 5
12 Time Management 1 2 3 4 5
13 Understanding the Community 1 2 3 4 5
14 Achievement of Desired Outcomes 1 2 3 4 5
15 OVERALL PERFORMANCE 1 2 3 4 5

Date: Signature of the Student

Page No: 37
 Evaluation by the Supervisor of the Intern Organization

Student Name: Registration No:

Term of Internship: From: To :

Date of Evaluation:

Organization Name & Address:

Name & Address of the Supervisor

with Mobile Number

Please rate the student’s performance in the following areas:

Please note that your evaluation shall be done independent of the Student’s self-
evaluation

Rating Scale: 1 is lowest and 5 is highest rank

1 Oral communication 1 2 3 4 5
2 Written communication 1 2 3 4 5
3 Proactiveness 1 2 3 4 5
4 Interaction ability with community 1 2 3 4 5
5 Positive Attitude 1 2 3 4 5
6 Self-confidence 1 2 3 4 5
7 Ability to learn 1 2 3 4 5
8 Work Plan and organization 1 2 3 4 5
9 Professionalism 1 2 3 4 5
10 Creativity 1 2 3 4 5
11 Quality of work done 1 2 3 4 5
12 Time Management 1 2 3 4 5
13 Understanding the Community 1 2 3 4 5
14 Achievement of Desired Outcomes 1 2 3 4 5
15 OVERALL PERFORMANCE 1 2 3 4 5

Date: Signature of the Supervisor

Page No: 38