Harsit Upadhya

Harsit Upadhya

AI Systems Engineer · M.S. Computer Science, Emory ’26

Open to AI engineering, developer tools, and data systems roles.

I build AI products, developer tools, and data systems people can use—from CartoVox's 300+ beta users to sub-second analytics across 385M events.

See selected work ↓ Get in touch
01

Research Interests

My research explores how large language models and deep learning can be applied to real-world problems in healthcare and science. Current focus areas: retrieval-augmented generation for domain-specific QA, medical image enhancement, graph neural networks, and digital health monitoring using passive technology interaction data.

Retrieval-Augmented Generation Large Language Models Medical Imaging Digital Health Graph Neural Networks Data Engineering Deep Learning
02

Selected Work

CodeNexus: AI Code Review Knowledge Graph

New

A topology-aware code review system that fuses deterministic static analysis with specialized LangGraph agents. Instead of dumping raw files into an LLM, CodeNexus builds an AST-driven project graph so agents can reason over imports, call chains, inheritance, security risks, and developer rationale with precise codebase context.

LangGraph Tree-sitter NetworkX Static Analysis Ruff Semgrep Bandit ESLint
GitHub
AST knowledge graph Local-first review Syntax / Logic / Security agents Terminal + JSON + PR comments

CartoVox: Fantasy Map Generator & Worldbuilding Studio

In Development

A local-first desktop world simulator for authors and game masters. CartoVox turns a written world brief into a reproducible spherical simulation—from plate tectonics, climate, rivers, and biomes to settlements, realms, routes, illustrated atlases, and a connected writing workspace.

Procedural Generation World Simulation Cartography Worldbuilding Local-First Desktop App
Case study Visit CartoVox Join Discord
CartoVox illustrated atlas showing named continents, seas, and routes across a generated world
49 measured map views 8 illustrated atlas sheets Brief-to-world seed search No account or cloud required 300+ beta users 300+ Discord members

One shared world record connects map exploration, atlas exports, settlements, lore, chronology, and manuscript tools.

ClickSim / EMBER: A Story-Driven Incremental Universe

Playable

Rebuild a dead universe from one surviving light. ClickSim, now titled EMBER, combines active clicking, idle growth, and progression across seven authored realms—each with its own economy, visual language, soundscape, and story.

Svelte 5 TypeScript Vite Canvas 2D Web Audio API
Play EMBER ↗ GitHub
EMBER's Emberlight realm with a glowing Heart, constellations, settlements, and Kindling upgrades
7 distinct realms 18 Kindlings per realm Local saves + offline progress No account or ads

E-Commerce Behavior Analytics Platform

Featured

A high-performance analytics platform that turns 385M+ raw e-commerce events into sub-second business intelligence. Built with a three-tier cloud-native stack (React + FastAPI + PostgreSQL star schema) on Google Cloud.

PostgreSQL 14 FastAPI React 18 Material-UI Google Cloud SQL Cloud Run Docker
Case study Project report GitHub
E-commerce analytics dashboard showing brand popularity metrics and a time-series chart of views, carts, and purchases
300–600× faster <1s queries 385M events 52GB dataset Star Schema

Monthly partitioning reduced scan size by 85% · 5 materialized views · B-tree indexing on product_id, user_session, event_type

Code Visualizer: Python Runtime Map

Interactive

A browser-based Python execution visualizer that makes runtime behavior spatial. Paste code, run it with Pyodide, and scrub through line-by-line traces of scopes, variables, references, loops, function calls, object mutations, stdout, and errors.

React TypeScript Pyodide Web Workers SVG Vitest
Case study Dashboard GitHub
Code Visualizer dashboard showing a highlighted Python line, array state, variables, and playback controls
sys.settrace snapshots Timeline playback Shareable #cv links

View all projects & research →

03

Publications

Ankit Kumar Upadhyay, Harsit Kumar Upadhya

XNLI 2.0: Improving XNLI Dataset and Performance on Cross-Lingual Understanding

IEEE 8th I2CT Conference · 2023

PDF arXiv

View all publications →

04

Experience

Jul 2026 – Present

Research Assistant (Part-Time)

Emory Temporary Services (ETS), Emory FIT Lab · Atlanta, GA

  • Continue collaborating with an Emory faculty mentor on research and technical work supporting the FIT Lab

Jan 2026 – May 2026

Graduate Teaching Assistant

Emory University · Atlanta, GA

  • Instruct 40 undergraduates in Data Science 100, covering R tidyverse, data cleaning, visualization, and EDA

Jan 2025 – May 2026

Graduate Research Assistant

Emory FIT Lab · Atlanta, GA

  • Built automated pipeline to extract and analyze Amazon Alexa voice interaction logs using Python/Selenium
  • Identified technology engagement patterns indicating functional decline in older adults for digital health monitoring

Oct 2024 – Dec 2025

Proctor

Emory University, Office of Undergraduate Education

  • Supervised 50+ exams while enforcing academic integrity protocols and supporting 300+ students during testing sessions

May 2025 – Oct 2025

VP, International Student Affairs

Graduate Student Government Association (GSGA)

  • Selected executive board member; served as primary liaison for 500+ international graduate students

Feb 2025 – Aug 2025

International Student Liaison

Emory University, Laney Graduate School

  • Mentored international graduate students through academic and cultural transitions, strengthening peer support networks and community-building initiatives
  • Collaborated with university staff to surface student challenges and improve orientation programming and campus integration
05

Get in Touch

Open to AI/ML engineering, data systems, software roles, and research-backed product work.

If the problem has messy data, real users, and a result worth measuring, I’d like to hear about it.

Or via GitHub or LinkedIn.