Nathan Lara

I build, evaluate & deploy AI products.

RAG pipelines, LLM agents, and the evaluation systems that prove they work — shipped for BCG X, the United Nations, and PwC.

Selected Work → Résumé Email

Currently

AI Product Builder — BCG X

Education

MS, Artificial Intelligence in Business — Gabelli School of Business (2026)

Focus

AI deployment · forward-deployed engineering · product · evaluation

Shipped for

BCG X · United Nations · PwC

Based

New York, NY

Links

LinkedIn GitHub

Selected Work

Five case studies

01 AI Blocks Open-source visual AI/ML programming environment — graphs compile to runnable Python. 1st place, vibeFORWARD. React · Python 02 M&A Deal Intelligence A deployed agentic pipeline for PwC monitoring 51 live sources with structured deal extraction and an analyst dashboard. GPT-4o · RAG 03 AEGIS Browser-native multimodal threat detection — motion-gated Claude Vision with a confirmation flow. 1st place, Grayscale. Claude Vision · WebRTC 04 UN-LEX Enterprise legal AI adopted by the UN Office of Legal Affairs — 9,000+ tribunal cases, cited, inside Microsoft Teams. Copilot Studio · Azure 05 AI Wealth Strategy Board-level strategy case for a RAG-grounded wealth assistant — $9.6M NPV across 10,000 Monte Carlo scenarios. Azure · Monte Carlo

About

I'm an AI product builder with an MS in Artificial Intelligence in Business from Fordham's Gabelli School. I build AI products at BCG X, and I've shipped production AI for the United Nations and PwC — tools that carry real workloads and get used by real people.

My work lives across the full deployment layer: designing RAG pipelines, wiring up LLM agents, and building the evaluation harnesses that separate a convincing demo from a system you can put in front of executives. I care about AI that actually works in production, not just in a slide.

My path into AI ran through finance and business — a BBA, a year directing investment research, and leading student orgs (I grew one council's membership 160% as CEO). That business-first instinct is why I build AI around the decision it's meant to support, not the other way around.