Abhishek Pundir

Worked as Reviewer at Mercor.

Build and review AI-focused software deliverables, focusing on code quality, testability and reproducible results.

• Design repository-backed coding-agent benchmarks with Docker environments, reference patches and executable tests.
• Analyse agent traces and review generated patches for correctness.

• Develop and validate terminal benchmarks through automated testing and code review.
• Reconstruct failing build environments to support benchmark validation.

Evaluate and refine STEM-focused LLM outputs to improve logical consistency and conceptual accuracy.
Review complex problem-solving responses for factual correctness and reasoning depth.
Collaborate with interdisciplinary teams to enhance AI-assisted learning and analytical tools.

Worked as AI Generalist Expert at Mercor.

Train and fine-tune language models for reasoning quality, instruction-following, and conversational accuracy.
Identify reasoning gaps and annotation inconsistencies, contributing to improved model alignment and coherence.
Support prompt optimization strategies to enhance contextual relevance and response reliability.

Worked as Human Data Manager at micro1.

Worked as Robotics Data Generalist and Quality Analyst at micro1.

Assisted in experimental setup management and technical documentation.
Supported data analysis and reporting for academic research activities.

Conducted large-scale reasoning evaluation and linguistic quality assessment for production LLMs.
Applied structured analytical frameworks to assess logic flow, factual accuracy, and instruction adherence.
Supported human-in-the-loop workflows for continuous model refinement.

Issued Jun 2026

Issued Mar 2026

Issued Jun 2026

Issued Dec 2025

A young graduate enthusiast to work hard and hone my skills.