Rohan Khatri

CS @ SJSU

Spotify Now Playing

GitHub Contributions

Education

San Jose State University

Bachelor of Science in Computer Science

Expected Graduation: December 2026

GPA: 3.94/4.0

Organizations & Activities

- AI & Machine Learning @ SJSU

Hobbies

Gym

- Lifting for 2 years

- Chasing 315 lbs bench

- Current bench PR: 275 lbs

Video Games

- Valorant peak rank: Immortal 3

- Clash Royale peak rank: Royal Champion

Trying New Foods

- Always down to expand my palette

- I like finding "underground" spots and gatekeeping them

Work Experience

Oct 2024 – Feb 2025 DataGlacier Data Scientist Intern
Jan 2025 – May 2025 AI & ML @ SJSU Lead Full-Stack Developer
May 2025 – Aug 2025 Maximus Software Engineer Intern
Aug 2025 – Present California DMV Software Engineer Intern

Projects

Forge

LangChain · React · TypeScript · Vite

  • Engineered a LangChain-based multi-agent debate engine that stress-tested ideas across adversarial and supportive personas before synthesizing execution-ready plans
  • Built a React/TypeScript frontend for structured prompt intake, live debate transcript rendering, and streamed final-plan output across a multi-stage agent workflow
  • Orchestrated parallel adversary/ally evaluation rounds and automated endpoint configuration, enabling faster deployment and more transparent decision refinement

PaperLink CalHacks 2025 Runner-Up

Next.js · FastAPI · Claude · Lava · Chroma · Fish Audio

  • Built a multimodal end-to-end pipeline using Lava to generate concise arXiv research summaries in under 1s
  • Integrated Fish Audio TTS for concurrent audio playback and synchronous text highlighting for vision-impaired users
  • Optimized MiniLM–Chroma vectoring with SQLite caching, reducing LLM calls 45% and enabling instant reloads

EVA

Python · FastAPI · Whisper · HuggingFace · Groq · ElevenLabs

  • Developed a real-time voice-to-voice AI coaching platform for guided self-reflection and supportive guidance
  • Engineered a FastAPI backend leveraging Whisper + FFmpeg for speech-to-text, HuggingFace Transformers for emotion classification, and Groq LLMs for context-aware responses, achieving 98%+ transcription accuracy
  • Automated text-to-speech synthesis with ElevenLabs API for adaptive, natural audio responses in live conversations

ZeroFootprint

React · FastAPI · Gmail API · PostgreSQL · Redis · Docker

  • Built an MCP-driven Gmail API pipeline to scan inboxes and draft privacy requests, minimizing user data exposure
  • Implemented Redis job queues and a PostgreSQL schema for re-scans, status tracking, and full audit logs

AI Discord Bot

Python · Meta-Llama · AWS EC2 · Flask · Discord API

  • Fine-tuned TinyLlama 1.1B model on 40k+ Discord messages, to replicate the user's conversation style
  • Deployed the model on AWS EC2 through Flask via Docker containerization, serving requests across 3 servers
  • Integrated data preprocessing workflows to convert 1,000+ messages to JSONL, filtering sensitive information

Tech Stack

Languages

  • Python
  • Java
  • JavaScript
  • TypeScript
  • SQL
  • HTML / CSS

Frameworks

  • React
  • Next.js
  • FastAPI
  • Flask
  • Spring Boot
  • Vite

AI / ML

  • PyTorch
  • Scikit-learn
  • HuggingFace
  • OpenCV
  • MediaPipe
  • LangChain
  • Whisper
  • Groq

Cloud / DevOps

  • AWS
  • EC2
  • Lambda
  • S3
  • Docker
  • Git
  • GitHub Actions
  • Splunk

Data

  • PostgreSQL
  • MongoDB
  • Redis
  • Pandas
  • Chroma
  • Dask