I am a seasoned Machine Learning Research Scientist/data scientist/Engineer leader with 17+ years of experience. My expertise is building and innovating statistical/machine learning models (Bayesian and Frequentist modeling techniques) to help businesses advance their data journey. Pushing the limits of how humans and AI reason through language—one commit at a time.
Oid Labs: https://www.oidlabs.com/
- Currently, in stealth. Experimenting with building safe, personalized assistants for Intelligent Process Automation using LLMs for vertical tasks.
- LeIA, Legal Intelligent Assistance
h2oai:
- Led research and developed services to improve ways to evaluate LLM models with Automated QA generation (task-specific synthetic test generation using auto prompt optimization) and alignment optimization (LLM robustness).
- Led efforts on LLM2SQL assistant for QnA on structured tabular data using SQL generation, https://github.com/h2oai/sql-sidekick
- Wave Image Styling Playground**: Creative generative AI exploration for interactive image styling.
- Responsible for research, concept, and production of Model-Analyzer, a unified interactive framework for simulation (What-If scenarios) and adversarial robustness, to continuously explore and evaluate predictive model behavior and limitations. https://docs.h2o.ai/wave-apps/h2o-model-analyzer/get-started/what-is-h2o-model-analyzer
- AutoInsights**: Collaborated closely with late Dr. Leland Wilkinson (Chief Scientist) to invent and launch an automated, self-service AI system that surfaces hidden data insights (auto-EDA) using natural-language narratives. https://docs.h2o.ai/wave-apps/h2o-autoinsights/get-started/what-is-h2o-autoinsights
- Responsible for driving ML innovation/product and business strategies to improve the model interpretation ideas in h2o Driverless AI MLI.
Before joining h2o.ai
- Datascience.com (Lead AI Research Scientist), bought by Oracle
- Led, researched, developed, and commercialized the need for Model Interpretation/Explanation for better model evaluation. Open-sourced our implementation for broader adoption of model explanation. Skater https://youtu.be/dmXF7WxT63E. We were among the first to offer interpretation as an enterprise solution.
- ...
Engagements: I am not very good at keeping track of these forms of engagements, but trying to get better at it. Sharing links to a few engagements from the past,
- Unpacking AI hype to reality with Kyle Polich: https://www.youtube.com/watch?v=x8ES6eHV43k
- Lexoid: https://oidlabs.com/blog/introducing-lexoid
- https://medium.com/the-story-within/state-of-text-to-sql-dc3e3e4f8c64
- SlideShare link: https://www.slideshare.net/pramitchoudhary
- https://www.youtube.com/watch?v=Mc_KSibrRVc
- https://opendatascience.com/model-evaluation-in-the-land-of-deep-learning/
- https://www.youtube.com/watch?v=fV9XFGN1BZ8
- Interviews:
- ...
📫 How to reach me: pchoudha@uci.edu; pramit@oidlabs.com