Methods and Systems for Automated Coding Assessment
US Patent Pending 63/296,939
Compressing a two-week study into an afternoon
What it takes to put an LLM in the middle of a regulated, document-heavy workflow — and where the model is the easy part.
Aayush
Nandkeolyar
I build AI-powered products and health-tech research platforms — and lead the teams that ship them.
I lead engineering teams working at the intersection of applied AI, real-time systems, and healthcare research — taking ambiguous problems from architecture through to production. I've built products from zero and scaled platforms people now depend on, and I care most about the decisions made early: the ones that quietly determine whether a system holds up as it grows.
Selected Work
-
SocialPlai
A crowd-powered machine learning platform that diagnoses ASD and ADHD in adolescents from digital social interactions — pairing a real-time multiplayer research environment with a human-in-the-loop labeling pipeline.
-
OverlineIQ
An LLM-driven system that turns weeks of expert document review into a few hours, spanning document ingestion, automated asset classification, and IRS-aligned report generation.
-
CHIME
A real-time music co-creation platform for autistic adolescents, designed to be administered by music therapists — built around latency budgets tight enough for people to play together in time.
-
MLPro
A LeetCode-style platform for AI/ML practitioners, with an integrated live technical interview environment and an educational portal — built from zero to launch.
Selected Writing
-
-
Engineering for research is a different job
Lessons from building NIH-funded platforms where the requirements are hypotheses, not specs.
-
Real-time is a product decision before it's a technical one
Notes from building WebRTC and WebSocket systems for interviews, therapy sessions, and multiplayer research tasks.