Your team already has data, workflows, and infrastructure in place — you don't need a rebuild, you need AI that fits into what's already working. That's where I come in.
I embed with teams to integrate LLMs and agentic systems into existing production environments, not standalone demos.
Recent results:
- Rebuilt a legacy MS system into a full-stack Django + Next.js application for a Canadian power utility, with zero data integrity loss
- Integrated an LLM into an existing communication workflow, cutting review time 15%
Built a fully local multi-agent system for internal HR/accounting queries — all inference on client infrastructure, zero external API calls, for sensitive data
- Delivered client-facing ML integration (auth + REST API) that led to a $500K contract
Fine-tuned 7–13B parameter models (LLaMA, Falcon) with 4–8 bit quantization, cutting compute cost 30%
Stack: Python, FastAPI/Django, Next.js, PyTorch/Transformers, vector search (Pinecone, ChromaDB), AWS/GCP/Azure.
8+ years shipping software, the last 2+ focused on deploying AI into environments that already exist. I learn your codebase and constraints first, then build — not the other way around.