Diyorbek Ibragimov
Senior Computer Science student
Carnegie Mellon University in Qatar
ibragimov at cmu.edu
LinkedIn
GitHub
Explore: About · Experience · Projects · Research · Fun
About
I am interested in low-level computer systems, especially operating systems, distributed systems, and systems for AI infrastructure.
Currently, I am seeking Summer 2027 internships / full-time roles / research opportunities.
Experience
Software Engineer Intern
- Building an agentic system that processes large policy corpora, identifies contradictions, and retrieves verified facts quickly.
Undergraduate Researcher
- Designed a three-phase allocation heuristic for multi-model ML inference across edge devices, achieving up to 5× higher throughput than Round Robin and within 10% of an MILP optimal solution, while reducing allocation time by two orders of magnitude.
- Developed a reactive autoscaler with accuracy-aware model selection, dynamic batching, and soft GPU partitioning to meet strict accuracy and latency SLOs, cutting request drop rate by 15%.
- Implemented a Python event-driven simulator modeling device heterogeneity, ML model inference, Poisson request arrivals, and queuing dynamics to evaluate the system across 15 inference tasks.
Open-Source Contributor
- Optimized issue-response by deploying an Azure-hosted FastAPI service that pushed triage alerts in Slack chats.
- Boosted ML accuracy 15% by collaborating with IBM to conduct data mining on 8K GitHub issues with MLflow.
- Improved debugging experience for 200+ contributors via Gemini 1.5 prompt-routing CI hooks in GitHub Actions.
Software Engineer
- Reduced Firestore query latency 20% by analyzing query patterns in BigQuery and refactoring server architecture.
- Enhanced price-quote accuracy 30% by optimizing database management and migrating 10K+ Firestore records.
- Initiated GCC-region expansion with a TypeScript microservice integrating DHL APIs for international shipping.
Projects
x86 Kernel
A preemptive, multithreaded x86 kernel built in C and x86 assembly.
- Built support for 20+ system calls, including fork, exec, wait, and vanish.
- Implemented virtual memory with per-process page tables, zero-fill-on-demand paging, and memory validation.
- Extended the kernel with a paravirtualized hypervisor supporting guest execution, hypercalls, and virtualized interrupts.
EduPro — LMS Platform
A learning management system for school staff and students.
- Maintained 99% server uptime for 1K monthly users by Dockerizing the Django platform on AWS EC2 with CI/CD.
- Coordinated with 5 schools to help 150+ teachers conduct remote exams via auto-grading and CSV analytics.
Offline AI First-Aid App
An iOS app that provides offline AI-based first-aid guidance.
- Led the launch of a first-aid app powered by 4-bit quantized Llama 3.2 for offline AI inference using MLC LLM.
- Designed an emergency RSS feed to alert local residents about disasters and evacuation points.
Uzbekistan Embassy Website
A landing page and content platform for the Uzbekistan Embassy in Qatar.
- Built an admin panel for staff to publish multilingual content with Django Rosetta.
- Supported domain registration, server hosting, and secure email communication.
Avis
A platform that connects ecological issues with customized maps and trustworthy charities.
- Combined customized Google Maps with local and global charity resources.
- Built a web interface for exploring critical ecological problems by location.
A*lbert
An arcade puzzle game inspired by Q*bert, built for a university arcade machine.
- Developed physics, level management, enemies, and sound effects.
- Awarded “Best Classic Award” by Professor Eduardo Feo Flushing.
Research
The Benefits and Risks of Artificial Intelligence: Decentralized AI Governance to the Rescue
During my senior year in high school, I wrote a research paper under the guidance of my Information Technology Teacher about the Ethics of Artificial Intelligence (AI), in which I argued for the development of the Democratic Model of AI Governance. In particular, I described the current state of AI, its unprecedented progress, and its potential to revolutionize every aspect of our individual and collective lives, which brings both advantages and disadvantages. To counter the negative consequences like rising unemployment and the AI arms race, I drew from the work of two Harvard researchers to argue for a three-layered model of AI Governance that actively involves the government, public and private institutions, and citizens across the world to contribute to the safe development and deployment of advanced AI models.
Fun
Leetcode 200 Questions in 9 hours
To celebrate the end of the summer semester at Carnegie Mellon University in Qatar, my friend and I decided to solve 200 Leetcode questions in one day. It took us 9 hours to complete this challenge. You can watch the stream here.