Python across the stack — APIs, RAG, and the data behind them.

ai / enterprise-rag

Agentic RAG for enterprise knowledge retrieval

Morgan Stanley engagement · 2025

Contributed to an OpenAI-based RAG system with an agentic, tool/function-calling workflow that routes queries to the correct retrieval source and returns grounded, citation-backed answers. Applied prompt engineering for constrained generation, and exposed retrieval and generation steps as Python service endpoints for downstream apps.

ai / rag-assistant

LangChain RAG assistant, built and shipped

Independent client work

Built a production-use RAG assistant for insurance agency clients: policy and coverage documents embedded into a vector store, with LangChain orchestrating multi-step retrieval and response chains. Served behind a FastAPI endpoint for natural-language policy queries, with Pytest coverage on retrieval relevance before rollout.

python / aircraft-configurator

Aircraft surface configurator

Libellule Monde · 2023 to now

A Django platform that analyses aircraft exterior regions and estimates the time, resources and cost of surface treatments. I led it from user interviews through to production, and own the PostgreSQL schema and query performance underneath it as the dataset has grown.

python / parts-labels

Parts label generation service

Libellule Monde · ongoing

A FastAPI service that generates enterprise-specific labels, stickers and plates for aircraft parts, with search across a large part-number database. Redis sits in front as a caching layer, which cut database round trips and made search suggestions fast enough to be useful while typing.

python / data-pipelines

ATA documentation ETL

Libellule Monde

Pipelines that pull structure out of ATA-standard aerospace documentation and load it into the analytics platform, so maintenance and repair teams can search and reason over it instead of opening PDFs. I own the query paths, indexes and storage strategy underneath, tuned against real usage as data volume grew.

python / trade-surveillance

Trading activity extraction

First Derivative · 2025

Python jobs that pull trader activity from NAS file systems, databases and Splunk and turn unstructured logs into structured JSON: timestamped queries, user actions, originating screen. The Risk team uses that feed for anomaly detection.

python / device-integration

Medical device data integration

Bacancy Technology · 2022 to 2023

FastAPI and Flask endpoints connecting Philips Biotel devices to a client portal, with secure storage and real-time processing of streaming device data. Django applications alongside it, hosted on EC2 with Nginx and Gunicorn.