AP · systems
Systems EngineerLow-latency · DistributedHyderabad, IN · CSE ’24–’28

Abhinav Pabbaraju

I work where correctness and speed fight each other — distributed consensus, compiler backends, GPU physics, order books. Chasing the fastest lap in software.

abby@systems:~$tail -fphalanx/raft.go
Focus
Systems·LL
Building
F1·2026
Shipped
14K+LOC
Next
LOB·µs

The short version.

Third-year CS & Engineering student who likes the hard, unglamorous parts of computing — the ones where a system has to stay correct when everything fails, and fast when it doesn’t.

StudyCSE · 2024–2028
DomainsDistributed · Compilers · GPU · HFT
StackGo · C++23 · Rust · TS · CUDA
CertsNVIDIA DLI · Harvard CS50

My work lives close to the metal: Raft consensus and linearizability checking, compiler backends emitting x86-64, GPU N-body physics, and the kind of tail-latency thinking trading systems demand. I care about numbers that don’t lie — p99s, instruction counts, throughput under chaos.

Off the clock I’m building an F1 analytics engine, turning CS papers I like into running code, and generally trying to make software that would survive a wet Sunday at Spa.

GoC++23RustTypeScriptPythonCUDAThree.jsNext.jsPrometheusKubernetesGSAPx86-64 · NASMRaft consensus
scroll to pan · click a card
keep scrolling

Come in.
We’re open late.

a Tokyo backstreet · 23:47 · still raining
深夜食堂SYSTEMS DINER写真
psst — the alley leads
to my field notes

the kitchen never closes · five dishes on tonight · counter seats only

scroll to walk the room · click a lit spine

On the grid this season.

Three lanes open at once: the F1 dashboard getting its final polish, a systems paper turning into running code, and the next benchmark on the bench.

Building

F1 2026 analytics

Polishing the live dashboard — real APIs, a Monte Carlo championship simulator, and a 3D globe, all in a single premium-feel app.

open it ↗
Reimplementing

Papers → code

Working through PagedAttention as a mini-vLLM build — the best way to actually read a systems paper is to make it run.

Sharpening

Systems & low-latency depth

Pushing toward FAANG and HFT — an LMAX-style limit-order-book matching engine is next on the bench, chasing microseconds.

Let’s build
something .

Open to internships and collaborations in systems, low-latency, and infrastructure. Fastest reply is by email. — no really, email me :)

^ donut.c, but it’s my website — spinning live, zero images, all math