Man Patel
Man Vijaybhai Patel
I've spent three years across research, consulting and industry. At De Vinci Research Center I built a recommendation system that proposes lower-carbon NoSQL data models from measured query patterns. That work became two first-author papers: a demo at WISE 2026, and a regular paper at IEEE Big Data 2026 that learns those costs so the ranking takes well under a second. Before that I audited models under the EU AI Act at Stellantis, measured the footprint of GenAI platforms at Wavestone, and built predictive models and ETL pipelines at GIFEC and Wordex Infotech.
Most of my work comes back to efficiency. A model that costs half as much to serve is often worth more than one that scores half a point higher. I also write about ML engineering here and make videos about it on YouTube, mostly to check that I really understand it.
Work Experience
Built a recommendation system within the DaMoOp & FACT-DM frameworks that generates optimised NoSQL data models from measured query patterns. Published as a first-author regular paper at IEEE Big Data 2026 (19.4% acceptance rate) and a demo paper at WISE 2026.
Modelled the three-way trade-off between query latency, cloud cost and carbon footprint to decide when a schema should migrate.
Designed benchmarking pipelines to measure system behaviour at scale and flag data models that have gone obsolete.
Measured the electricity, water and carbon footprint of GenAI platforms to inform clients' Cloud FinOps and Green IT strategies.
Translated client business requirements into technical advisory deliverables.
Ran AI risk assessments aligned with the EU AI Act, deployed a risk assessment tool, and audited models for bias and fairness.
Built automated KPI dashboards tracking the lifecycle of AI projects across Stellantis business units.
Built a location-aware predictive model suite (KNN, Random Forest, Decision Trees) reaching 70% accuracy under 5-fold cross-validation, then presented the results to over 200 business leaders.
Engineered production SQL ETL workflows that raised data accuracy by 20%, plus Power BI dashboards the client used daily.
Education
MEng, Data and Artificial Intelligence · Aug 2024 – Aug 2026 · Graduated
Bachelor's degree, Computer Applications · Aug 2019 – Aug 2022
Publications
Instead of simulating every candidate data model, DaMoOp* learns the time, money and carbon cost of each query from one simulator run. It ranks 135 candidate models in about 0.6 seconds instead of 8, and picks the same cheapest models as the simulator. With Jihane Mali and Nicolas Travers. 203 of 1,046 submissions were accepted.
A tool that proposes optimised NoSQL data models from measured query patterns, weighing query latency, cloud cost and carbon footprint.
Recognition
Top 8 finalist at a hackathon run by ESILV, Capgemini, Renault Group and Dassault Systèmes; pitched to their executives.
IBM watsonx.data · IBM Generative AI
English C1 (IELTS) · French B1 · Hindi native · Gujarati
Let's work together
I'm looking for a full-time ML Engineer or Data Scientist role in Paris or remote, and I can start now. Email is the fastest way to reach me.