Nikash Chhadia

I build AI agent systems and the infrastructure that trains them. Currently a Member of Technical Staff at Mirendil. Previously, I worked on Grok at xAI, built real-time energy forecasting for Tesla's Powerwall and Megapack fleet, and built agentic validation for quantum hardware at AWS. I studied CS & Math at Stanford.

Work

  1. 2026 –

    Mirendil Member of Technical Staff

    RSI

  2. 2025 – 26

    xAI Member of Technical Staff

    Grok

  3. 2025

    Tesla Machine Learning Engineer Intern

    Real-time ML forecasting for Opticaster, the energy optimization engine across 1M+ Powerwall and Megapack sites. Built a simulation data pipeline, first launched at Fremont Factory, saving $4M+/year in energy costs.

  4. 2025

    AWS Braket Software Development Engineer Intern

    Built an agentic AI system to verify QPU calibration metrics, and automated quantum device onboarding (95% less time, no manual error).

  5. 2025

    Arc Institute Research Engineer

    ML systems and agent orchestration for generative biology: research, infra and experimentation.

  6. 2024

    AWS EC2 Software Development Engineer Intern

    Features for the EC2 provisioning system (5,000+ RPS), and a debugging redesign that cut the logs and sites needed per host by 80%.

Education

  1. MS

    Stanford University Computer Science

  2. BS

    Stanford University Mathematics and Computer Science

    Tau Beta Pi.

Research & Projects

  • Structural Attention in OCR Backbones for LaTeX Paper

    Fine-tuned OCR backbones for image-to-LaTeX on a procedural pipeline rendering 50,000+ expressions/sec under diverse noise and corruption, and surfaced attention heads that track long-range delimiter structure.

    Stanford CS231n · 2024

  • Covert Video Steganography via Deep Learning

    An end-to-end 3D CNN that hides data in video under adversarial conditions. Adaptive loss weighting improved message secrecy by 48% and capacity by 8% over baselines.

    Deep learning · 2024

  • Automated Options Wheel Trader

    A fully automated, delta-based Wheel strategy on live options-chain data with real-time execution. Tuned put and call sizing, rolling and exits to a Sharpe ratio above 2.

    Quant · 2025

  • Voice-Controlled Computer-Use Agent GitHub

    Speak a task and an LLM breaks it into steps, reads screenshots to find its targets and drives the mouse and keyboard.

    TreeHacks '24

  • Citadel Datathon GitHub

    One of 75 selected from thousands of applicants. Analyzed what moves airline stocks: flight reliability, jet fuel prices, events and natural disasters.

    Citadel Securities · 2023

  • Heap Allocator

    Implicit and explicit heap allocators that simulate dynamic memory allocation in C.

    Systems · C

Say hi

Always happy to chat about interesting ideas. chhadia@stanford.edu