I build AI that
holds up in production.

RAG systems, AI agents, and backend infrastructure — for teams who need it to still work after the demo.

About

3 years building AI systems for regulated industries — RAG pipelines, fine-tuned LLMs, and backend infrastructure that runs at 99.9% uptime.

I care about what happens after launch. Auditable, cost-aware, and boring in the best way — if it can't survive a compliance review, I don't ship it.

3+ Years Experience

80% Manual Effort Reduction

What I actually build with, day to day.

Llama, JAIS, Qwen, GPT, Claude, Gemini, Hugging Face Transformers, vLLM, PEFT (LoRA, QLoRA).

RAG, LangChain, LangGraph, Hybrid Search (Dense + BM25 + RRF), Cross-Encoder Reranking, Text-to-SQL.

Python, SQL, FastAPI, WebSockets, PostgreSQL, DynamoDB, Redis, ChromaDB, Vespa.

AWS, Docker, Kubernetes, MLFlow, Weights & Biases, CI/CD, Model Monitoring.

Same four steps, every project. No surprises.

30 minutes, no pitch. We talk through the problem and whether I'm the right fit.

A fixed-scope proposal — timeline, deliverables, and price. No surprises later.

Weekly demos of real, working progress — not status updates.

Tested, documented, and handed off so your team can own it.

Jun 2023 — Dec 2025

Deployed NLP pipelines for financial and legal document automation on AWS, processing hundreds of documents daily at 99.9% uptime.
Cut manual document review by 80% (financial) and 60% (legal) through fine-tuned classification and extraction systems.
Fine-tuned BERT and LegalBERT with PEFT (LoRA), improving classification F1 from 0.70 to 0.85+.
Architected hybrid RAG pipelines (dense + BM25 + RRF + cross-encoder reranking) for regulated document intelligence.
Built session-aware agentic systems on Redis and DynamoDB for persistent, multi-turn legal analysis workflows.