Open to AI/ML roles · US work authorized · No sponsorship required
Vikhyat Chauhan

Vikhyat Chauhan

AI/ML & AI-Coding Engineer

I build AI systems and the tooling around them — from agentic coding harnesses to production RAG.

I work end-to-end on agentic and AI systems — LangGraph multi-agent pipelines, retrieval built from scratch, and local-LLM inference. Two years shipping production software at GE HealthCare, then an M.S. at Virginia Tech defending a thesis on brain-inspired autonomous navigation. I build the unglamorous parts well, so they hold up in the real world.

ex-GE HealthCare M.S. Virginia Tech · 4.0 GPA NSF-funded research US Work Authorized
Download résumé GitHub LinkedIn Email
2024 — 2026

Graduate AI/ML Researcher

· Virginia Tech — Blacksburg, VA
  • Designed and experimentally validated the Conflict Architecture (CA) framework for brain-inspired UAV navigation (NSF #2204780, PI: J.M. Paul), achieving a simultaneous 34.6% reduction in compute energy and elapsed time over baseline through parallel conflict resolution (p < 0.0001, n=1,000 runs across 3 real-world-mapped environments).
2022 — 2024

Software Engineer II

· GE HealthCare — Bengaluru, IN
  • Led the end-to-end modernization of a legacy MRI image-stitching application serving GE’s 1,000+ system global fleet, re-architecting into containerized microservices that increased throughput from single-process to 20+ parallel operations while meeting the 99% clinical accuracy threshold. Sole SWE II in the organization to receive the Impact Award.
  • Segregated a core C++ algorithm from a legacy monolith and designed a microservices architecture (TypeScript, Java REST, Docker/Helm, RabbitMQ) plus a CI/CD pipeline, adopted as the standard across multiple applications.
  • Scaled the team from 1 to 4 engineers and partnered with global platform architects and clinical specialists to deliver validation, regulatory sign-off, and onsite technician training.
2020 — 2022

Co-Founder & Sole Technical Lead

· The Next Move Electronics (TNM) — Delhi, IN
  • Founded and bootstrapped (~$10K self-funded) a home-automation hardware startup, owning every layer from PCB design and firmware to cloud backend and a Flutter/Alexa mobile app.
  • Owned the entire customer relationship as sole founder — discovery, onboarding, and support across 100+ households — converting early prospects into first design-win customers.
  • Architected an AWS microservice backend (EC2, Lambda, RabbitMQ, MongoDB, MQTT), scaling from 0 to 1,000+ devices at 99.9% uptime and cutting field bug reports 98% through a structured hardware-software validation workflow.
2018 — 2020

Research Intern

· Mahindra Susten — Pune, IN
  • Built a Linux-based autonomous quadcopter controller (MAVLink + advanced control) and integrated it into Mahindra Susten’s solar ops—automating thermal fault detection and reporting, delivering a 35% reduction in labor costs and faster issue triage.
Personal · Full-stack · Live

Hearth — Shared-Household App + NL Action Layer

A full-stack collaborative household app — recurring chores with two-way Google Calendar sync, shared lists, and Splitwise-style expenses — plus a natural-language command bar where Claude tool-calls map straight onto the app’s zod mutation schemas (validate → Drizzle), guarded by a clarify-or-decline policy and an offline eval harness.

Next.js ·TypeScript ·React ·Supabase ·Drizzle ORM ·Vercel
Live → View source →
2-way
Calendar sync
zod-validated
NL → mutations
Vercel
Deployed
Personal · Healthcare AI · Local-first

Specced — Clinical Notes → FHIR

A QLoRA-fine-tuned Qwen2.5-Coder-7B runs a LangGraph plan → retrieve → act → evaluate loop that turns free-text clinical notes into de-identified, terminology-coded, schema-valid FHIR R4 — every output validated and field-scored against gold. Local-first by design: PHI never leaves the box.

Qwen2.5-Coder ·QLoRA ·LangGraph ·FHIR R4 ·LanceDB ·Unsloth
View source →
0.42 → 1.00
Resource F1
20/20
Held-out pass
1.00
De-id recall
Personal · Production

Professional RAG Pipeline

Two-stage retrieval (BGE-base → MS-MARCO cross-encoder) with SHA-256 semantic cache and per-query token tracking. Built from scratch — no LangChain — and deployed as a containerized service.

Python ·FastAPI ·BGE ·Cross-Encoder ·Cloud Run ·Docker
Read the deep-dive → View source →
−30%
API overhead
50+
Eval queries
LLM-judge ↑
Faithfulness
Virginia Tech · M.S. Thesis

CANavigator — Brain-Inspired Drone Navigation Framework

Conflict-resolution framework on a MISD architecture for 3D autonomous UAV navigation. Validated across three real-world-mapped simulation environments with statistically significant gains.

Python ·MISD ·Control Systems ·Simulation
View source →
−34.5%
Energy
−34.6%
Time
p < 0.0001
Significance
Engineering

Three changes that made my coding agents reliable

I spent weeks building a multi-agent harness for software delivery. The biggest surprise: as the models get better, the leverage moves out of the model and into the system around it. Three changes moved my reliability more than any prompt tweak.

Read →
Research

A drone that holds three minds in parallel — and only picks one when forced to

Most autonomous systems force a single resolved output. Biology doesn't. CANavigator runs three navigation algorithms in parallel and only commits when the environment demands a response — and it beats the fastest single algorithm on both time and energy simultaneously.

Read →
Aug 2024 — May 2026

M.S., Computer Engineering (Thesis Track)

· Virginia Polytechnic Institute and State University — Blacksburg, VA
  • Defended M.S. Thesis on Brain-Inspired Drone Navigation System using Conflict Architecture — 4.0 GPA.
  • Coursework: CS5424 Advanced Machine Learning, CS5465 Applications of Machine Learning, ECE5504 Computer Architecture.
Jul 2016 — Dec 2020

B.E., Electronics and Communication Engineering

· Uttarakhand Technical University — Uttarakhand, IN
  • Graduated with First Division.
2026

Brain-Inspired Drone Navigation Using the Conflict Architecture

V. Chauhan

M.S. Thesis · Virginia Tech

2025

Multi-Agent Stock Prediction Systems: Machine Learning Models, Simulations, and Real-Time Trading Strategies

D. Dave, G. Sawhney, V. Chauhan

arXiv:2502.15853

2021

Master-Slave Microcontroller Communication System for Home Automation

V. Chauhan (co-inventor), The Next Move Electronics

Granted Patent · Patent No. 202111060973

GE Impact Award

Recognized for independently modernizing a critical legacy MR imaging application to production microservices, directly improving clinical software delivery at scale.

NSF-Funded Graduate Research Assistant

Supported on NSF Award #2204780 (CCF: Small: Paradox and Brain-inspired Computer Architecture, PI: JoAnn M. Paul); contributed to thesis research on brain-inspired autonomous systems.

Honor Society of Phi Kappa Phi

Inducted, Virginia Tech Chapter (2026); top 10% of Master’s graduate students across all disciplines.

Omicron Delta Kappa National Leadership Honor Society

Selected for membership, Alpha Omicron Circle, Virginia Tech (2026 Cohort); recognizes achievement in academics, research, and service.

I'm focused on production AI / ML engineering roles — inference optimization, agentic systems, and RAG pipelines that have to actually ship. Background in FDA-regulated medical AI is a bonus for clinical-grade deploys. US-authorized, no sponsorship needed, open to relocation.

Email vikhyat.chauhan@gmail.com
Phone +1 (201) 616-6373
LinkedIn vikhyat-chauhan
GitHub Vikhyat-Chauhan
Location Blacksburg, VA · Open to relocation