Summary
Mayank Agarwal is a senior research engineer and machine learning specialist with 12 years of experience, currently focused on post-training foundation model work and agentic capabilities at Liquid AI after a productive tenure at IBM Research. He led design and deployment of tool-calling stacks and outcome-based reward models (ToolRM) used for synthetic data filtering, rejection-sampling fine-tuning, and RL policy training, and has trained models up to 70B parameters with SFT/DPO/GRPO. His research spans federated learning, controllable code generation, and fairness tooling—work that earned multiple IBM Research Accomplishment awards and publications at NeurIPS and ICML. Based in Cambridge, MA, he combines deep research rigor with production engineering, integrating reward-model pipelines and benchmarks like Nestful into real systems. An interesting through-line in his career is turning research prototypes (e.g., execution-validated synthetic data pipelines and CLAI agent integrations) into practical, audited frameworks used across teams.
12 years of coding experience
11 years of employment as a software developer
Senior Secondary, Mathematics and Computer Science, Senior Secondary, Mathematics and Computer Science at St. Peters College, Agra
Master of Science (M.S.), Computer Science, Master of Science (M.S.), Computer Science at University of Massachusetts Amherst
Bachelor of Technology (B.Tech.), Electrical and Electronics Engineering, Bachelor of Technology (B.Tech.), Electrical and Electronics Engineering at Visvesvaraya National Institute of Technology
English, Hindi