Skip to content
View umangPokhriyall's full-sized avatar
2 followers · 8 following

Achievements

Achievement: Pull SharkAchievement: YOLO

Achievements

Achievement: Pull SharkAchievement: YOLO

Highlights

  • Pro

Block or report umangPokhriyall

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
umangPokhriyall/README.md

Umang Pokhriyal

Systems engineer focused on Rust, distributed systems, low-latency infrastructure, and applied cryptography.

I build projects to answer concrete engineering questions, measure the results, and publish both the successes and the failures. Every performance or security claim links to the benchmark, experiment, or test that produced it.

I'm self-taught, and this profile is my public engineering notebook. Everything here is intended to be cloned, reproduced, and inspected.

Website & long-form writeups: https://umangpokhriyall.github.io

Repo The one result Source
frost-ed25519-kit Built an RFC 9591 FROST threshold signer after demonstrating a practical ROS forgery against the original implementation. Includes differential testing, fuzzing, and documented threat model. legacy/results/ros_forgery.txt
low-latency-lob Explores data-structure tradeoffs in matching engines. A flat-array order book that excelled on synthetic workloads became 288× slower than a BTreeMap under real BTCUSDT traffic, with hardware counters explaining why. bench/results/throughput.csv
Rust-Tcp-Server Compares 11 server architectures built on a shared sans-IO core. Shows that reducing syscall count alone does not guarantee higher throughput under C10K workloads. bench/results/c10k_summary.csv
Coingate Implements exactly-once withdrawal processing over an at-least-once event stream and validates it with exhaustive crash-injection testing. chaos/results/summary.md
proctor Studies scheduler latency, verification cost, and distributed coordination, including a measured analysis of Redis round-trip overhead and accompanying security model. bench/results/metal-m4-large/sched/

Measured on rented AMD EPYC bare metal where the claim depends on hardware; the crypto and exactly-once results are hardware-independent and reproduce anywhere. Each repo says which is which.

Open to systems and infrastructure roles.

Pinned Loading

  1. frost-ed25519-kit frost-ed25519-kit Public

    Hand-rolled FROST-Ed25519 threshold signatures in Rust. RFC 9591, sans-IO, no unsafe. Includes a working forgery of the naive scheme it replaced.

    Rust

  2. low-latency-lob low-latency-lob Public

    Low-latency limit-order-book engine in Rust. Four book structures benchmarked to the cache line, with the microarchitecture teardown of why the "optimal" one loses on real data.

    Rust

  3. Rust-Tcp-Server Rust-Tcp-Server Public

    Every TCP server concurrency model in Rust, from accept-loop to io_uring, behind one sans-IO core and benchmarked to C10K.

    Rust

  4. Coingate Coingate Public

    Exactly-once processing of an at-least-once stream, proven by a chaos harness that crashes at every statement boundary. Rust, 0 double-sends at READ COMMITTED.

    Rust

  5. proctor proctor Public

    A zero-trust control plane for untrusted compute in Rust: probabilistic verification, in-memory shard-scoped crypto, and a backpressure-aware scheduler, all measured.

    Rust