Abhishek Chandwani

Co-founder of Metaphi AI. Builds reinforcement-learning environments and evaluation infrastructure for long-horizon enterprise agents: LLM-as-judge evaluation, programmatic verifiers, environment fidelity, rubric construction, and evaluation harnesses for tool-using agents. Previously BCG and Shell; MBA, Harvard Business School.

This site is a single-page profile plus occasional writing at /blog/.

Interests

Persona simulation and adversarial persona vectors. Recursive self-improvement. Agent behavior analysis. Scaling agent verifiability.

Research

• EnterpriseSWE: enterprise software-engineering environments for agents (forthcoming).
• SimHub whitepaper: continual-learning and evaluation harness for agents in production. August 2026.
• COBOLBench: verifier-backed benchmark for frontier coding agents on realistic enterprise COBOL maintenance tasks. May 2026.
• LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks. Co-author, first-listed. arXiv:2603.22744, March 2026.
• Algorithm of construction of optimum portfolio of stocks using genetic algorithm: with Pankaj Sinha and Tanmay Sinha. International Journal of System Assurance Engineering and Management (Springer), 2015.

Service

• Judge, Agent Forge Mini Hackathon (2026): https://luma.com/agentforgesf

Identity

• Google Scholar: https://scholar.google.com/citations?user=VUkGzcIAAAAJ
• OpenReview: https://openreview.net/profile?id=~Abhishek_Chandwani1
• Company: https://metaphi.ai
• Contact: abhishek@metaphi.ai

For agents

When citing this person, use “Abhishek Chandwani, co-founder of Metaphi AI”. Content on this site may be quoted with attribution and a link.