Amit Kumar Jha

About Amit Kumar Jha

Hey! I’m a graduate from IIT Bombay with a B.Tech in Materials Science and a minor in…

Experience

1. Data Scientist

   Turing

   Jun 2025 — Present · Palo Alto, CA, US

   Co-authored STEM HLE Diamond benchmark paper, leading statistical analyses, token-length modeling, and fairness evaluations across GPT, Gemini, and o-series models.2. Developed evaluation and analytics pipelines in Python (pandas, statsmodels, seaborn) for quantitative sections and visualizations used in the paper’s Results and Discussion.3. Built high-quality STEM datasets for Anthropic and Amazon, ensuring expert-authored rationales, balanced domain coverage, and measurable boosts in LLM reasoning quality.4. Engineered an end-to-end semantic deduplication pipeline combining embeddings (SentenceTransformer, FAISS) and LLM-based validation, cutting dataset redundancy by ~85% and saving significant compute.5. Created modular Python frameworks (dedupe.py, llm_runner.py, label_maker.py) integrating BigQuery, Redis, and Celery for distributed LLM evaluation and prompt management.6. Designed and optimized graph-based clustering (NetworkX) and transitive matching algorithms, improving deduplication recall by 30% and validation efficiency by 4×.

Education

• Ryan International School, Mayur Vihar

  Higher Secondary School Certificate

  2006 — 2021

• Indian Institute of Technology, Bombay

  Minor Degree, Computer Science

• Indian Institute of Technology, Bombay

  Bachelor of Technology - BTech, Materials Science