Stoneset.ai
Stoneset classifies AI systems against the EU AI Act — deterministically, with every decision traceable to a specific article, paragraph, and annex. Built for Series A/B AI startups navigating compliance.
Co-Founder & Data Scientist
Stoneset classifies AI systems against the EU AI Act — deterministically, with every decision traceable to a specific article, paragraph, and annex. Built for Series A/B AI startups navigating compliance.
Fine-tuned an LLM at Roche for multi-cancer early detection, deployed as a POC. Authored a research abstract on AI evaluation approach for MCED.
An independent case study analysing Tesla's market position, business strategy, and competitive landscape using data-driven insights — covering vehicle segments, manufacturing scale, and the broader EV industry trajectory.
Comparative machine learning study on real NASA spacecraft telemetry — benchmarking classical, deep learning, and Transformer-based models against expert-labelled anomaly sequences, deployed as a mission-control style dashboard.
I spent my placement year at Roche building AI tools that actually got used — OCR automation, a fine-tuned LLM, data pipelines that replaced manual processes people had been living with for years.
Now I'm co-founding Stoneset.ai, a compliance tool for the EU AI Act. I run product and operations. We're early and building fast.
Stoneset.ai
Co-founded
In progress
Research publication
Stoneset.ai
Deterministic EU AI Act compliance — classifying AI systems by risk tier with every conclusion traceable to a specific article. Responsible for product and operations.
Microsoft
Selected for an 8-week programme with professionals across Microsoft, covering technology, leadership, and strategy.
Roche
Built OCR automation reducing processing time from hours to under three minutes. Improved data retrieval by ~80%. Designed and deployed a fine-tuned LLM on HuggingFace. Configured CI/CD pipelines. Worked across teams in Switzerland, Canada, China, and India.
University of Westminster
Machine Learning, Data Engineering, Software Development, Statistics, Cybersecurity. Built supervised ML models for cancer patient mortality prediction — over 85% accuracy.
Have an idea worth building? I’m open to partnerships, collaborations, and interesting problems. Whether you want to work together or just compare notes — my inbox is always open.
© 2026 Rohan Chaudhary. All rights reserved.