Skip to content

Applied Mathematics @ UC Berkeley

Embedded ML · robotics & autonomy · applied AI

I build AI systems, robotics platforms, and embedded intelligence — from FPGA-accelerated machine learning to assistive mobility robots. I'm studying Applied Mathematics at UC Berkeley, research embedded ML at Santa Clara University's WIN Lab, and filed a provisional patent for BALANCE, a hybrid legged-and-wheeled mobility device.

At a glance

Patent
BALANCE · 63/743,085
Paper
Youth Innov. Journal '25
Lab
SCU WIN Lab
Robotics
VEX Worlds qualifier
Based
Berkeley, CA

FIG. 01sense · infer · actuate

analog indigital out
01

Featured work

01Research · ongoing

SCU Wireless Intelligent Networks (WIN) Lab · Jul 2025 – present

FPGA-Accelerated Machine Learning

FPGA-based acceleration for machine learning and wireless-communication algorithms — high-throughput, low-latency signal processing and channel estimation.

  • FPGA
  • Embedded ML
  • Signal Processing
  • Channel Estimation
  • Low-Latency Systems
02Provisional patent · Published

Co-developed with a student team · Youth Innovation Journal, Fall 2025

BALANCE — Assistive Mobility Robotics

A hybrid mobility device — wheeled on flat ground, robotic-leg assistance on stairs and uneven terrain. Provisional patent No. 63/743,085.

  • Assistive Robotics
  • Mechatronics
  • LiDAR
  • Navigation
  • CAD
03Internship

Rainier Labs · Sunnyvale, CA · Jun – Aug 2025

Robotics Hardware & Embedded Systems

Expressive robotic head system — display integration, CAD redesign for hardware fit, and audio component evaluation.

  • Embedded Systems
  • Robotics Hardware
  • CAD
  • Human-Robot Interaction
02

Technical focus areas

  • 01

    Embedded ML

    PyTorch/TensorFlow models, quantized and ported to FPGAs and constrained hardware

  • 02

    Robotics & autonomy

    Embedded and hardware robotics, autonomous systems

  • 03

    Applied AI systems

    CarePath, Klarity VoiceNote, GuestFlow — shipped AI product prototypes

  • 04

    Voice agents

    Grok Voice, Retell AI, ElevenLabs across four shipped prototypes

  • 05

    Human-assistive robotics

    Mobility and accessibility systems

03

How I build

Less a process than a bias. The domain changes; how I approach a problem doesn't.

Frame

Find the real limit

What actually binds this —compute, data, time, trust

Reduce

Smallest thing that proves it

One question per prototypeCut scope, not rigor

Contact

Test against reality

Real inputs, real usersBenchmarks come second

Iterate

Keep what survives

Discard what didn't earn itThen go again

04

Recognition

  • AIME qualifier (American Invitational Mathematics Examination)

    2026

  • Eagle Scout — Boy Scouts of America

    2025

05

Resume

Current · PDF

Experience, research, and selected technical work.

OpenDownload
06

Collaborate

Open to research collaborations, internships, and technical projects.

tarun_yadgirkar@berkeley.edu