Nikhil’s Pursuit
GitHub

NIKHIL KADAPALA / PHD STUDENT · UNH

What does the agent actually know?

I build AI systems that have to work on messy, real-world text, then I try to measure whether they actually do. My PhD is on agent evals: knowledge bases, retrieval, and memory.

Read the research ↗Browse systems →
research map4 threads · 4 artifacts
KNOWLEDGE BASESRETRIEVALMEMORYEVALSCheckThat!CLEF 2025 · paperagentic-ragmultimodal RAG + evalsHarnessBoxagent harness primitivesResAlign AIcareer intelligence
hover a node to trace its links
NOWPhD · University of New Hampshire
FOCUSAgent evals · memory · RAG
PAPERCheckThat! · CLEF 2025

01 / THE ARGUMENT

Retrieval is not knowing.

A high score can describe a system nobody can use. I keep running into the same gap: the metric goes up, and the thing a person actually needed still isn’t there. That distance — between a plausible answer and a useful one — is most of what I work on.

How I approach it

PUBLISHED RESEARCH

The score is not the work.

Research index
UNH at CheckThat! 2025: Fine-tuning Vs Prompting in Claim Extraction
We compared fine-tuning and prompting setups for claim extraction. FLAN-T5 won on METEOR, while iterative self-refinement sometimes produced claims a fact-checker would actually want to work with. That gap is the whole plot: evaluation has to stay connected to the user and the work the system is meant to support.
2025 · Claim extraction · Evaluation · CLEF 2025
Paper ↗Writeup ↗Read the writeup

SELECTED SYSTEMS

Things I actually shipped.

All systems
ResAlign AI
A career copilot for fit, preparedness, and alignment—not another apply-to-everything button.
Product / career intelligence
01 · In the wildOpen
HarnessBox
Sandbox and harness primitives for coding agents. Zero dependencies, on purpose.
Open source / agent infrastructure
02 · BuildingOpen
agentic-rag
Multimodal agentic RAG with a real evaluation harness. “It felt pretty good” is not a metric.
Research engineering
03 · Open sourceOpen
CheckThat
Noisy social posts to concise, checkable claims—class project to paper to pip install.
Paper → package
04 · PublishedOpen

CURRENTLY INTO

Open questions.

[ 01—03 ]
01 / Knowledge & memory

Knowledge & memory

What an agent actually knows, not just what it retrieved once. Retrieval is a lookup; knowing is what survives the next turn.

02 / Agent evals

Agent evals

Score the behavior against intent, not a storyboarded workflow. A golden trace is a useful fixture and a terrible definition of success.

03 / Inference engineering

Inference engineering

Disaggregated prefill/decode, speculative decoding, prefix caching. Cheaper and faster serving, without pretending that's the same as a smarter model.