Ankit Singh Chauhan

Ankit Singh Chauhan

AI Researcher — Large Language Models (agents, RAG, PEFT/QLoRA), Causal Inference and Probes, Knowledge Graphs

Indianapolis, IN · Indiana University, Indianapolis

Resume GitHub LinkedIn Email

About

I am an AI Research Engineer at Indiana University Indianapolis, working under the guidance of Dr. Jeremy Price. I am currently designing agentic AI systems—specifically multi-model agent architectures with model routing, tool-use, retrieval, and prompt orchestration on serverless platforms using Python and TypeScript.

Prior to that, I worked as a Research Assistant under Dr. Sunandan Chakraborty on the NSF-funded CATpc project, curating instruction-tuning datasets and fine-tuning Llama-3.2 with QLoRA and DPO for a pedagogically aligned educational AI tutor. With Dr. Chakraborty I also completed an independent study, CultureEval, using PCA over ~97k survey respondents to quantify cultural alignment in three LLM families. During this time I contributed to two papers: Benchmarking LLMs for Pairwise Causal Discovery in Biomedical and Multi-Domain Contexts at IEEE Big Data 2025, and Mod-Guide: An LLM-based Content Moderation Feedback System at ACM COMPASS 2026 (Accepted).

What I work on

  • LLM fine-tuning — LoRA / QLoRA / DPO on Mistral-7B and Llama-3.2. Lifted Triple-F1 on structured extraction from 0.32 → 0.78; instruction-tuning datasets curated to Cohen’s κ 0.88 inter-annotator agreement.
  • Knowledge-graph extraction — two-pass extraction pipelines on Amazon Bedrock Agents (Lambda, DynamoDB), routing documents via a Haiku classifier between Sonnet and a fine-tuned Mistral-7B at ~20× lower per-token cost. React + Cytoscape.js HITL review UI on top.
  • Production ML & MLOps — FastAPI services, Docker/Kubernetes, CI/CD, deployed across AWS / GCP / Azure.

LLM-Driven Community Asset Knowledge Graph (Mistral-7B + Bedrock)

· AI Researcher · Indiana University Indianapolis

Two-pass LLM extraction pipeline on Amazon Bedrock Agents — Haiku classifier routes documents between Sonnet and a LoRA-fine-tuned Mistral-7B — lifting structured extraction from 0.32 → 0.78 Triple-F1 over 8.3k records, with a React/Cytoscape.js/FastAPI HITL review UI validated at 0.82 Cohen's κ.

  • LLM
  • LoRA
  • Mistral-7B
  • RAG
  • Amazon Bedrock
  • Lambda
  • DynamoDB
  • FastAPI
  • React
  • Cytoscape.js
  • Neo4j

CultureEval: Quantifying Cultural Alignment in LLMs

· Academic Project · Indiana University Indianapolis

PCA-based evaluation framework over ~97k survey respondents and 96 sociocultural indicators, surfacing systematic underestimation of religious-traditional values in Llama-2, Gemma-3, and Phi-4 (Cohen's d 0.89–1.17).

  • LLM evaluation
  • PCA
  • Llama-2
  • Gemma-3
  • Phi-4
  • Cohen's d

Pedagogical Instruction-Tuning for Llama-3.2 — NSF CATpc

· Research Assistant · Indiana University Indianapolis (NSF-funded)

Built a 7.2k high-quality instruction-tuning dataset (Cohen's κ 0.88) and fine-tuned Llama-3.2 with QLoRA + DPO for the NSF-funded CATpc project — +14% pedagogical alignment, −8% hallucinations vs. baseline.

  • LLM fine-tuning
  • QLoRA
  • DPO
  • Llama-3.2
  • dataset construction
  • RLHF
Code NSF Award

See all projects →

Selected Publications

See all publications →