Full Stack AI Engineer at Alsonotify, a Digibranders brand · Mumbai
I build AI products and the systems around them: retrieval, agents, APIs, and the interfaces people use. I care about what happens after the first successful demo—whether answers are grounded, actions are controlled, and the product holds up when something fails.
OyeChats · AI product engineering
My contributions include fixing a relevance gate that excluded useful knowledge-base content, consolidating duplicated RAG pipelines, and moving qualification work onto a durable job queue. The prompt consolidation documented in the remediation PR reduced one assembled prompt from 15,233 to 8,120 tokens.
Retrieval diagnosis · Prompt, streaming & durability work
CleanStart · Product performance and correctness
I batched CMS relationship lookups and fixed cache invalidation that could run before a database write committed. For the documented 25-row list, the change reduces roughly 75 relationship requests to 2.
CMS performance · Cache correctness
|
A document-aware assistant built around tools, policy decisions, approval controls, and execution traces. Making an agent's actions inspectable is part of the product. AI orchestration · Retrieval · Tool policies |
A desktop companion with a developer's working day: commits, test runs, deploys, and Claude Code sessions become little moments of character. Built primarily for macOS. Developer tools · Native UI · Playful interaction |
|
A rental marketplace that goes beyond listings: vendor onboarding, rental workflows, checkout, payment verification, disputes, and administration. Full-stack product · Marketplace workflows · Payments |
Find images with a sentence or another image. CLIP embeddings and vector search connect a FastAPI backend to a React interface. CLIP · ChromaDB · FastAPI · React |
More builds: GitHub Mirror, a GitHub-footprint explorer with deterministic visual signatures, and explainable phishing detection with feature selection, SHAP, and LIME.
AI & data — Python, FastAPI, LangChain, CLIP, ChromaDB, scikit-learn, PostgreSQL, Redis.
Product — TypeScript, React, Next.js, Node.js, Tailwind CSS.
Delivery & experiments — Docker, GitHub Actions, AWS, MLflow, DVC.
I also send focused fixes upstream:
- llm-gateway — preserve Anthropic server-tool parameters when forwarding requests.
- Readest — make multi-device file sync converge.
- Pandoc — restore the missing bar on a blockquote's first line in ANSI output.
More contributions and project details →
Background & earlier work
- Full Stack AI Engineer, Alsonotify / Digibranders — April 2026–present.
- Data Science Intern, Code B Solutions — February–April 2025.
- Master in Data Science and Analytics with AI, IT Vedant with IBM — June 2024–June 2025.
- B.E. Computer Engineering, University of Mumbai — 2024.
- Published research: A Framework to Make Voting System Transparent Using Blockchain Technology, IJREAM, April 2024.
Earlier projects include computer vision for attendance, a CNN classification workflow with MLflow and DVC, and an email assistant with human escalation.
Building something useful with AI? Let's talk.
The intro shows a real training run on synthetic data. How the animation is made.