From physics notebooks
to production ML.
I came to AI from an unusual angle — a BSc in Physics, Chemistry and Mathematics from Bengaluru North University, then a leap to Paris for a Master's in AI & Data Science at Aivancity. The science background gave me a habit I lean on every day: be skeptical of your own results, run the controls, check whether what you measured is what you think you measured.
Since landing in Paris in 2024, I've worked across surprisingly different domains. Multimodal retrieval at NeuralTeks. Medical signal processing for seizure detection at AURA. Document intelligence research at Aivancity. NLP for recommendations at IntHealth. Currently agentic RAG and forecasting at INSEAD in Fontainebleau.
Alongside the day job, I do research on clinical machine learning and LLM reliability. That means a sole-authored preprint on LLM-augmented reinforcement learning, two book chapters in production with Elsevier and CRC Press, and presentations at IANLP in Casablanca, the Health Data Hub in Paris and PyConES in Barcelona, with a poster accepted at WiML @ NeurIPS. I'm now looking for full-time roles applying multimodal foundation models to real problems.
The thread is the same in every job: take messy real-world data, build the right model for it, and make it useful to someone outside the notebook.
I write clean Python, take MLOps seriously (Docker, MLflow, version-everything), and care about whether what I built actually works in production. Outside of paid work, I volunteered with Linkee and Secours Populaire distributing food across Paris, and designed UX for MoralityCode, an ethical-AI app, as a UN Volunteer with the Action Lab for Development.
Particulars
Research
Education
Stack
Currently learning
Multimodal foundation models, agentic systems, and the practical side of evaluating how reliable LLM-driven applications really are. Also: French, slowly.