I'm an early-career AI & backend engineer who builds close to real systems — search pipelines, agentic workflows, and scalable services that need to be reliable, efficient, and explainable.
- 🛰️ Currently shipping a production social platform solo at a stealth startup (Next.js + Supabase + AWS)
- 🔬 Interested in: retrieval & ranking systems · agentic AI orchestration · backend performance at scale
- 🎯 Open to: AI Engineer / Backend Engineer / Software Engineer roles
| Platform | Rating | Rank |
|---|---|---|
| 🍴 CodeChef | 2483 — 6★ | #7 in India · #85 Globally |
| ⚡ LeetCode | 2591 — Guardian | Top 0.13% Globally |
| 🔵 Codeforces | 2225 | Master |
Languages
AI / ML
Backend & Full Stack
Cloud & DevOps
Full Stack Engineering Intern — Stealth Startup · Jan 2026 – Present
Solo-built a production social platform from scratch. Reduced DB calls ~50% and API payload ~80% via request path optimization on Next.js + Supabase. Shipped real-time auth, Supabase Realtime subscriptions, and a zero-downtime CI/CD pipeline on AWS EC2 + PM2.
AI & Frontend Development Intern — Edunet Foundation · Aug – Oct 2025
Built a Whisper + T5 + DistilBART pipeline converting 60+ min lectures to structured notes, flashcards, and MCQs — cutting inference time ~80% via GPU caching. Also shipped a responsive React portfolio with lazy loading.
Web Development & SEO Intern — Badakarobar · Mar – Jun 2025
Improved Lighthouse scores ~45% and automated sitemap generation entirely via Python scripting.
|
Autonomous Supply Chain Risk Monitor 3-agent system (Watchman → Analyst → Dispatcher) for 24/7 risk monitoring. Self-correcting confidence loop expands evidence scope when certainty is low. Persistent deduplication suppresses duplicate alerts. Deployed on Cloud Run with sub-3s cold start.
|
Hybrid RAG Engine with Caching FAISS (semantic) + BM25 (keyword) hybrid retrieval. SQLite content-hash caching cuts indexing latency ~90% on repeated runs. Non-blocking AsyncIO FastAPI backend. Transparent per-result Vector vs. Keyword score breakdowns.
|
|
Skin Lesion Classification with XAI EfficientNet-B0 on HAM10000 — 81% val accuracy, 0.67 macro-F1 across 7 imbalanced classes. AMP + Albumentations cut GPU time 2.8× and memory 40%. Grad-CAM visualizations for clinical interpretability at <50ms/image.
|
Full-Stack Campus Social Platform Deployed and used by 200+ students across campus. Modular Express controllers + indexed MongoDB queries improved backend efficiency ~35%. JWT auth, Cloudinary CDN, Tailwind UI with real-time status updates.
|
Active contributor to 9 production-grade repos — 300k+ combined ⭐
| Repo | Contribution |
|---|---|
| google/adk-python | Fixed multilingual LLM-as-a-Judge evaluation for Hindi, Chinese, Arabic (PR #3503) |
| mem0ai/mem0 | Azure AI Search vector store integration via factory pattern (PR #3549) |
| HackerRank Hiring Agent | Patched memory leak in PDF processing; fixed Windows UTF-8 encoding failures (PR #133, #129) |
pandas 43k ⭐ |
3 merged PRs — Ruff B905 zip compliance + PyArrow DateTime regression tests |
Matplotlib 20k ⭐ |
Overhauled TwoSlopeNorm colorbar docs with practical examples (PR #30639) |
TheAlgorithms/Python 208k ⭐ |
Hopcroft-Karp bipartite matching O(E√V) (PR #13233) |
SQLFluff 8k ⭐ |
Extended SQL parser for PostgreSQL OPERATOR() syntax (PR #7163) |
|
|
|
|
|
|
Always open to interesting problems — reach out.