ML Engineer · AI Architect · Agentic AI Builder
📍 Bangalore, India
I build AI that drives real business outcomes — not demos, not POCs. I work across domains — from marketing intelligence and finance to consumer apps — wherever there's a hard problem that AI can genuinely solve. Specialising in RAG Systems · Multi-Agent Workflows · Predictive ML · Full-Stack AI Apps.
AI-native by design: I compress 6-month roadmaps into 6-week deliveries using LLMs, vector search, and agentic frameworks.
🔧 Open Source Contributor: statsmodels · SciPy · Microsoft AutoGen · LangChain · featuretools
📩 If you're hiring for Senior ML / AI Architect / GenAI roles — I ship fast, think in systems, and make AI work for your business. Let's talk.
PRs I've raised to production ML/AI libraries:
| Project | PR | What I fixed |
|---|---|---|
| 📊 statsmodels | #9876 | Improved docstrings across module |
| 🔧 featuretools | #2773 | Added 812 tests validating shapes across all 203 primitives |
| 🔬 scipy | #25535 | Added to stats, signal, integrate, optimize |
| 🤖 AutoGen (Microsoft) | #7897 | Fixed missing __init__ in agentic memory utils |
| 🦜 LangChain | #38567 | Fixed path traversal check rejecting valid directory names |
| 🦜 LangChain | #38568 | Fixed Chroma returning raw distances instead of normalized relevance scores |
Languages & Core
ML / AI
Full-Stack & Mobile
MLOps & Cloud
Self-hostable A/B experiment platform for comparing AI/ML models on real customer-support traffic. Runs controlled experiments between any two models — local sklearn .pkl artifacts, Ollama, OpenAI, or Anthropic — streaming live metrics via SSE. Welch's t-test confidence scoring, CSAT tracking, and a one-click Approve/Reject workflow that automatically promotes the winner to Production. Full data pipeline: ingest → preprocess → train → verify → register. Zero frontend dependencies — plain HTML/CSS/JS SPA.
Python FastAPI SQLite scikit-learn SSE Docker A/B Testing Statistical Inference
End-to-end MLOps pipeline for B2B lead conversion prediction. XGBoost + Optuna hyperparameter search, SHAP explainability, Evidently AI drift monitoring, FastAPI inference service, multi-stage Docker build, and GitHub Actions CI/CD to AWS ECS. ROC-AUC 0.950 on the UCI Bank Marketing dataset.
Python XGBoost Optuna SHAP FastAPI Docker AWS ECS GitHub Actions
🍽️ RecipeSnap
AI-powered iOS recipe manager built with React Native and Expo. Import recipes from URLs, photos, or video links — GPT-4o-mini extracts and structures everything automatically. AI remix adapts any dish to your dietary needs. Offline-first with optional Supabase cloud sync.
React Native TypeScript Expo OpenAI GPT-4o-mini Supabase Node.js
Real-world transaction anomaly detection for financial data — isolation forests, statistical baselines, and threshold tuning for low false-positive rates.
Python scikit-learn Pandas
NLP pipeline for extracting structured fields from unstructured invoice text — line items, totals, vendor info — using regex and transformer-based NER.
Python NLP Transformers
Check out my writing at BitsBytes →
"The goal is to turn data into information, and information into insight." — Carly Fiorina
⭐ If you find my work useful, consider starring my repos!