Built an end-to-end AI-driven investment research platform, transforming multimodal financial documents across 600+ companies into a structured, queryable knowledge base using RAG retrieval and LLM reasoning. Designed a production-grade retrieval system combining OCR (Mistral OCR), vector search (Weaviate), and hybrid retrieval strategies to improve contextual relevance and answer quality. Developed a multi-agent research and reasoning framework (FastAPI + OpenAI + Gemini) for generating structured investment insights from diverse sources. Implemented knowledge grounding with sentence-level citations to increase trust and reduce hallucinations in production. Reduced retrieval latency by ~50% via caching layers (MongoDB + vector cache), concurrency optimizations, and a pub-sub cache invalidation architecture. Built automated evaluation framework for internal AI agents and LLM workflows. Constructed data ingestion pipelines on GCP (Cloud Run Jobs, GCS, Docker) to continuously index large

Python Engineer at J.P. Morgan Global Commodities

September 1, 2022 - March 1, 2024

Led migration of a legacy production trading system tracking 100k+ deal flows; built an end-to-end migration framework and coordinated releases with global engineering, trading, and quant teams. Migrated a 10,000+ line Python 2 codebase to Python 3, improving maintainability and reliability of critical commodities trading infrastructure. Improved observability by building Splunk dashboards tracking system performance, latency, and deal-booking metrics across trading workflows. Collaborated with teams across London, New York, Buenos Aires, and Singapore to support global trading systems.

Machine Learning Engineer Internship at J.P. Morgan Corporate Technology

June 1, 2021 - August 1, 2021

Increased the observability of a production NLP-based country risk analysis application by building custom metrics with Prometheus, aggregating data with PromQL, and creating real-time visualizations using Grafana, resulting in enhanced monitoring and faster issue resolution. Implemented an automated synthetic data generation mechanism to support daily integration testing, ensuring robust validation of data pipelines and faster deployment cycles.

Machine Learning Engineer Internship at TokenMill

June 1, 2020 - September 1, 2020

Worked as part of a four-person engineering team at TokenMill, a startup specializing in automated text generation, document parsing, and NLP solutions. Extracted and classified sentences from more than 10,000 news sources to feed automatic sentence generators, ensuring high-quality, diverse data for downstream NLP tasks. Increased spaCy NER model accuracy from 85% to 95% by extracting custom entities from the dictionary and refining entity recognition, leading to more precise information extraction in production systems.