Mehrdad
Hessar
Senior Software Engineer at NVIDIA. PhD from UW. From internet of things to language models. I build systems that ship.
Senior Software Engineer at NVIDIA. PhD from UW. From internet of things to language models. I build systems that ship.
01 — About
I hold a PhD from the University of Washington, where I worked on platforms for low-power wireless systems, backscatter communication, and IoT at the edge.
Today I’m a Senior Software Engineer at NVIDIA, working on TensorRT for NVIDIA’s deep learning accelerator. Before that, I was Staff Engineer at OctoAI, leading LLM serving, Stable Diffusion fine-tuning, and model deployment infrastructure.
As an Apache TVM Committer, I maintained microTVM to bring ML compilation to bare-metal microcontrollers.
9
Peer-reviewed
publications
1
Distinguished
Paper Award
5+
Years in
industry
4
Side projects
live & shipping
02 — Industry
2024 — Present
NVIDIA
Working on deep learning accelerator compilers. Contributing to TensorRT backend
2021 — 2024
OctoAI (Acquired by NVIDIA)
Promoted from Senior to Staff SWE. Led text generation services (Mistral, Llama), fine-tuning for Stable Diffusion and LLMs, OctoAI CLI, and FastAPI server frameworks. Built caching systems enabling seamless continuous LLM deployment.
Summer 2020
Added TensorFlow Lite support in Chromium and analyzed on-device ML model performance on Android.
Spring 2020
OctoAI
Added Apache TVM compiler support for Azure Sphere and improved runtime performance with auto-tuned kernels.
Summer 2018
Microsoft Research
Built IoN, a long-range VHF IoT network with >20x LoRa coverage.
2017 — 2021
University of Washington
Dissertation: “Platforms and Technologies for Low-Power Wireless Systems.”
Published in NSDI, MobiCom, and Ubicomp.
2015 — 2017
University of Washington
2011 — 2015
Amirkabir University of Technology
03 — Research
2020
NSDI 2020
2017
Ubicomp 2017 — Distinguished Paper Award
As seen in: The Economist · IEEE Spectrum · Tech Review
04 — Side Projects
Beyond work: tools for developers, productivity utilities, and the occasional card game.
VS Code Extension
Open any ONNX file as an interactive neural network graph inside VS Code or Cursor. Zero setup, fully private, runs where your editor runs.
Multiplayer Card Game
A fast-paced online bluffing card game for 2–7 players. Run a jungle fruit shop, ring the bell, and outsmart the angry gorilla.
Chrome Extension
Import ICS and VCS calendar files to Google Calendar with drag & drop. Privacy-first — all processing happens locally, nothing leaves your browser.
Open-Source Hardware
The first battery-powered standalone software-defined radio built for IoT endpoints. Programmable over the air and idles at 30 µW — 10,000× lower than existing SDR platforms.
05 — Updates
Jul 2026
Launched NetronLink — VS Code and Cursor extension for interactive ONNX neural network visualization.
Mar 2026
Launched Durian Panic — a real-time multiplayer bluffing card game for 2–7 players.
Jan 2026
Released the iCalendar Chrome Extension — drag-and-drop ICS and VCS imports into Google Calendar.
Sep 2024
Joined NVIDIA as Senior Software Engineer — TensorRT automotive AI and safety certification.
2023
Attended tinyML Summit 2023.
Sep 2022
Promoted to Staff Software Engineer at OctoAI.
2022
Submitted MLPerftiny v1.0 results using microTVM.
2022
Attended tinyML Summit 2022.
Nov 2021
Became Apache TVM Committer.
Mar 2021
Joined OctoAI.
Jan 2021
PhD defended at University of Washington — thesis.
2020
Presented TinySDR at NSDI 2020.
2019
Presented NetScatter at NSDI 2019.
2018
Presented Towards Battery-Free HD Video Streaming at NSDI 2018.
2017
LoRa Backscatter wins Distinguished Paper Award at Ubicomp 2017.
2017
Presented the Long-Range Backscatter poster at Microsoft Research Summit.
2016
Awarded the Madrona Prize for backscatter research at UW Industrial Affiliates Day.
2016
Presented the On-Body poster at MobiCom 2016.
06 — Contact
Interested in collaboration, open-source, or just want to talk systems and AI? My inbox is open.
