Emaan Heidari Homepage ©
About
I'm a computer engineering + CS student at USC and I build high-throughput + low-latency systems software.
I'm interested in AI infrastructure, especially runtimes, training infrastructure, inference optimization, and model serving. I recently implemented a Flash-Decoding attention kernel with paged KV caches to study efficient LLM inference. I've also built a FPGA inference accelerator in Verilog, using fixed-point arithmetic, LUT activations, and quantization-aware training.
I've worked on edge systems across drone autonomy, rocket engine data acquisition, and AI accelerator firmware, and this summer I'll be writing embedded Rust as a SWE intern at Tesla.
Some things I've built
Drag around this tetrahedron. Try to count the number of faces.
- FPGA neural inference accelerator: pipelined Verilog,
8.95µs latency per inference, 95.1% accuracy with quantization-aware training.
- MurmurMatch: a college social platform.
Grew to 55k users (51% of Dartmouth, 27% of Stanford).
GCP autoscaling, indexed SQL with Redis, ~50ms p99.
Work experience
- Tesla: incoming SWE intern, Energy (Summer 2026)
- USC Liquid Propulsion Lab: C++ data acquisition for a liquid bipropellant engine + feed system
- SiFly: drone autonomy systems, PX4 failsafes, simulation framework
- TandemLaunch: firmware for an analog AI accelerator
Random
Things I like
- card/board games, roller coasters, on-euclidean geometry, ocean, fire, hiking, gym, reading, thinking
Favorite Quotes
- "All of humanity's problems stem from man's inability to sit quietly in a room alone." -Blaise Pascal
- "Premature optimization is the root of all evil" - Donald Knuth
- "Be so good they can't ignore you" - Steve Martin
Last fiddled: May 2026.