Selected work
Two AI products I build and run outside my day job, on one shared platform, and the Amazon work I can describe in public. The product case studies cover the problem, what I built, the tradeoffs, what went wrong and what I’d change at ten times the size; the Amazon one keeps to what I have published. Every fact links to a public source.
- Case studies
- 3
- Sources
- 11 public pages, cited per section
Case studies
-
Clarity Today
Outside work · I build and run it
A daily news product where language models write every story, every claim carries its source, and a story that can’t back its own citations doesn’t publish.
-
RetailIntel
Outside work · I build and run it
Retail intelligence for people making consequential retail decisions in India. A chain of AI agents reads 79 publications and files cited signals, each with a forecast and a counter-case.
-
Amazon Music: fraud and content moderation
2021 – 2022 · Senior Lead Software Engineer, Amazon Music
The team that built automated policy-violation detection for the Music catalog and a streaming fraud platform. Told from what I have published, without Amazon’s internal detail.
Earlier at Amazon
Semantic search on the product detail page
2017 – 2020 · Software Development Engineer, Amazon.in
On the Product Detail Page Platform team in Amazon.in’s retail organisation: semantic search on the detail page with deep-learning models and approximate nearest-neighbour lookups, then an offer-aggregation platform behind Home, Search and the detail page.
Source: LinkedIn: experience
Amazon work is described as I published it on LinkedIn: titles, team scope and what we built, with no internal numbers. Views are my own, not Amazon’s.