Skip to content
View burhanuddin6's full-sized avatar

Highlights

  • Pro

Block or report burhanuddin6

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.

Content in all repositories owned by your account will be closed.
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
burhanuddin6/README.md

Burhanuddin Aliasghar Ezzi

Backend engineer. I work on the compute layer of academic planning software — the part that has to be correct and fast at the same time.

Most of what I've built comes back to questions like these: how to improve the process?, why is this slow?, and what's the smallest change that makes it fast?
Designing a stochastic algorithm, A GPU path tracer, a genetic algorithm on CUDA, and a degree-audit engine in Django turn out to have more in common than they look.


Currently

  • Working on the audit engine at Stellic — the system that computes where a student actually stands against their degree requirements
  • Taking latency out of audit-computation APIs, mostly by finding the work that never needed to happen
  • Modeling planning and compliance rules.
  • Day to day: Python · Django · Postgres · Celery· ElasticSearch · pytest

Selected work

Corpora Generation for Urdu Grammatical Error Correction Findings of ACL 2026

Urdu grammatical error correction was stuck for a boring reason: there was almost no annotated data to train on. We mined real mistakes from the edit history of Urdu Wikipedia, learned the patterns behind them, and injected similar errors back into correct sentences — synthesizing aligned incorrect/correct pairs at scale. That's UrduGEC-Synthetic. We paired it with UrduGEC-Gold, a human benchmark built from real student exam papers, then fine-tuned models on the synthetic data and evaluated against the gold set to show the generated corpus actually holds up.

Pipeline code: gec_pipeline

accelerated_small_pt — A path tracer rebuilt on Taichi so it runs on the GPU. Small enough to hold entirely in your head, which makes it a good place to learn where parallel speedups actually come from.

IMGA_cuda — Island Model Genetic Algorithm on CUDA. Populations evolve in parallel across islands and migrate between them; the interesting part is the migration topology, not the fitness function.

ats — An applicant tracking system. Where I do product-shaped work instead of compute-shaped work, and remember that most software problems are people problems.

I write things down occasionally at burhanuddin6.github.io/blogsite.

By the numbers

Followers Stars Focus


Blog · ORCID · Open to talking about performance work, constraint solvers, and anything that makes a slow system fast.

Pinned Loading

  1. gec_pipeline gec_pipeline Public

    Jupyter Notebook

  2. ats ats Public

    Applicant tracking system

    JavaScript 2 1

  3. matt77hias/smallpt matt77hias/smallpt Public

    ☀️ The Rosetta smallpt (small path tracing) project

    107 9