Senior AI & Machine Learning Engineer specializing in neuromorphic computing, automotive vision, and reinforcement learning.
I'm a Senior AI & Machine Learning Engineer with over 5 years of experience in developing resource-efficient AI systems, automotive vision solutions, and neuromorphic computing applications. I optimize deep learning models for low-precision hardware and lead vision system deployments. Passionate about spiking neural networks, computer vision, and mentoring, I drive innovation in AI-driven solutions.
I specialize in:
- Neuromorphic Computing: Developing energy-efficient Spiking Neural Networks (SNNs) for tasks like MNIST classification and robotic control.
- Computer Vision: Building real-time vision systems for automotive applications using YOLO and transformers.
- Reinforcement Learning: Designing multi-agent systems and policy gradient methods for industrial automation.
- Mentorship: Guiding teams in AI model optimization and deployment.
Explore my work in the snn_implementations directory, including:
- SSNN: Supervised SNNs for classification tasks.
- RSNN: Reinforcement learning with SNNs.
- Neuron Models: Evaluation of SNN neuron models.
| Project | Performance Gain |
|---|---|
| Smart Vision Systems | |
| Multi Agent DRL | |
| Accent Analyzer | |
| Neuron Models |
| Project | Description | Tech Stack | Impact |
|---|---|---|---|
| Smart Vision Systems | End-to-end AI pipelines for automotive vision using YOLO-World and Swin-Transformer. | PyTorch, Kafka, Docker | Enabled real-time detection on edge GPUs. |
| Multi-Agent DRL | Transformer-based multi-agent reinforcement learning for industrial agents at TRUMPF. | PyTorch, CUDA, ROS | 20% faster task completion. |
| Accent Analyzer | Streamlit-based tool for English accent detection in videos. | Streamlit, Azure Speech, FFmpeg | High-accuracy multi-platform support. |
| Neuron Models | Framework for evaluating SNN neuron models with backpropagation and surrogate gradients. | PyTorch, SNNTorch, Matplotlib | 85% accuracy on MNIST with optimized neuron models. |
Open-Source Contributions:
- Contributed to Huggingface Transformers (3 PRs merged).
- Maintained OpenCV-Python tutorials.
- Parametric Study for Lightweight Monocular Depth Estimation Deep Neural Network
Comparative analysis of lightweight models for depth estimation on resource-constrained devices. Link - Policy Learning with Spiking Neural Networks for a Robot Manipulation Task
Explored energy-efficient SNNs for robotic manipulation in neuromorphic environments. Link
- IEEE SB TUC Hackathon, 3rd Place (Sep 2023)
Developed EEG-based emotion decoding using generative models. - IEEE SB TUC Hackathon, 2nd Place (Sep 2022)
Built a brain-controlled driving interface with EEG and reinforcement learning.
- Programming: Python, C++
- Frameworks & Libraries: PyTorch, TensorFlow, Keras, OpenCV, Huggingface Transformers, SNNTorch, PyTorch Lightning
- Tools: Docker, CUDA, Git, ROS, Pandas, MLflow, Apache Kafka, NVIDIA Jetson
- Domains: Deep Learning, Computer Vision, NLP, Generative AI, Neuromorphic Computing, Multi-Agent Systems, EEG Signal Processing
- Robotics & Middleware: ROS, PyBullet, Brain-Computer Interfaces
Certifications:
- Deep Learning Nanodegree (Udacity, 2021)
- Azure Certified Machine Learning (2023)
Click to View Stats
Click to View Activity Graph
- LinkedIn: Osama
- Blog: Medium
- Email: [email protected]
- GitHub: Novalis133
Feel free to explore my repositories and connect to discuss AI-driven innovations!