Muhammad Sohail

Data & AI Engineer · Paris, France

Data & AI Engineer building production retrieval systems and self-hosted LLM infrastructure. I care about software that ships and keeps running.

About

I'm a Data & AI Engineer based near Paris. Most of my work lives in one corner of the field: getting real documents into a language model and trustworthy answers back out — retrieval-augmented generation — and building the infrastructure that keeps it running every day.

I run the internal AI platform at Université d'Évry Paris-Saclay, where it serves the HR and IT teams. I built the first version of it during my master's thesis, then stayed on to turn the prototype into something people actually depend on.

I lean toward self-hosted, open-source models when the data shouldn't leave the building, and toward boring, reliable infrastructure over whatever framework launched last week. I'd rather ship a system that handles auth, logging, and failure than demo one that only works on the happy path.

Experience

Data & AI Engineer · Université d'Évry Paris-Saclay Sep 2025 — Present

I design and run the university's internal AI platform for the HR and IT departments — from the retrieval pipeline down to the monitoring.
Built a multimodal document pipeline that ingests the messy PDFs people actually have, and a hybrid search that mixes semantic and keyword matching.
Operate self-hosted LLM serving with French and English support, plus custom integrations for directory lookups, ticketing, and workflow automation.
Set up logging and observability so I catch problems before users report them.

PythonLangChainFastAPIPostgreSQLFAISSDockervLLM

Data Science Intern — Master's Thesis · Université d'Évry Mar — Jun 2025

Built the first production HR & IT chatbot using retrieval-augmented generation, deliberately on open-source models for data sovereignty.
Engineered the document-parsing pipeline with Unstructured.io, Docling, and LlamaParse.
Deployed it inside the university data center — no data leaving the premises.

Data Analyst Intern · CodeClause (Remote) Oct 2023 — Apr 2024

Wrote ETL routines to clean and reshape messy real-world datasets.
Built interactive dashboards and statistical reports for non-technical stakeholders.

Selected projects