Harris Ahmad
PhD student in distributed systems · building concurrency & transaction systems · previously shipped production backends
Seeking Software Engineering / Research Internships for Spring / Summer 2027.
Email me about internships or referrals: harrisah@buffalo.edu
News
Selected for Y Combinator AI Startup School
Selected for YC AI Startup School 2026 in San Francisco (Chase Center) from a pool of ~30,000 applicants. Spent the program in office hours and 1:1s with founders and operators on AI products and systems.
About Me
I'm a PhD student in Computer Science at the University at Buffalo (SUNY), advised by Dr. Haonan Lu. I work on systems that make collaborative AI workloads reliable at scale — spanning transactions, distributed databases, and serverless backends.
Before starting my PhD, I spent 2+ years as a professional software engineer building production backends with FastAPI, MongoDB, Redis, and cloud platforms — including media APIs, SSO, and real-time systems at companies such as Linq.io and Its IT Group.
During my undergraduate studies at LUMS, I was advised by Dr. Zafar Ayyub Qazi, Dr. Ihsan Ayyub Qazi, and Dr. Mian Muhammad Awais. I conducted research on internet affordability and YouTube ad costs, publishing at ACM WebConf 2024.
Teaching: TA for Modern Networking Concepts (CSE 489/589) at UB, and previously for OOP in C++ and Distributed Systems at LUMS.
Research
I'm building transactional support for microservices, distributed databases, and serverless systems to enable collaborative AI. Current work (unpublished, manuscripts in preparation) includes:
- Evaluating concurrency and replication protocols (primary-backup, chain replication, two-phase locking) for real-time collaborative AI workloads on CloudLab
- Designing a distributed transaction reordering and coordination layer on systems such as FoundationDB and CockroachDB to reduce lock contention, aborts, and tail latency
- Broader agenda: making cross-service and serverless transactions practical for multi-agent / collaborative AI systems
This research is supported by NSF-funded projects at UB.
Skills
Core: Python, Go, C/C++, TypeScript/JavaScript, SQL · FastAPI / Flask / Node · Docker, Linux, AWS · MongoDB, CockroachDB, FoundationDB, Redis · distributed systems & transactions
Featured Projects
deadpush
Always-on guardian for AI coding agents — catches secrets, agent debris, and dangerous writes before they land in your repo.
- Problem: Long-running agents leak keys, commit scratchpads, and pollute context while you're away.
- Built: Real-time filesystem daemon, quarantine, git hooks, MCP proxy, and hardened / sandbox / CI enforcement tiers.
pip install deadpush
ChatLiberate
Open-source ChatGPT exporter that works where Settings → Export does not — including Business & Teams accounts.
- Problem: Business/Teams users can't use official export; third-party tools drop branches and images.
- Built: Chrome extension + Node CLI that emit official
conversations.json, full conversation trees, and attachments.
npx chatliberate -o ./my-backup
gitpull
Go CLI for multi-repo Git workflows — clone, sync, status, branch, and local AI helpers without context-switching.
- Problem: Managing many clones means repetitive pull/status/commit across directories.
- Built: Parallel clone/sync, workspace-wide status/diff/stash, plus optional Ollama-powered commit/standup/ask.
brew tap harris-ahmad/tap && brew install gitpull
Awaaz E Sehat
Flask + AWS Lambda eHealth platform: patient workflows, medical transcription, and a serverless data pipeline for 50K+ recordings.
Work Experience
Graduate Researcher
- Building a distributed transaction reordering and coordination layer on FoundationDB and CockroachDB in Go that reduces aborts, lock contention, and tail latency (work submitted to ACM SOSP 2026)
- Prototyped primary-backup, chain replication, and distributed two-phase locking in Go to stress-test protocol behavior for collaborative AI workloads on CloudLab
YC AI Startup School
- Selected participant at YC AI Startup School 2026 (Chase Center) from a pool of ~30,000 applicants — office hours and 1:1s with founders on AI products and systems
Software Engineer
- Shipped an on-demand media-generation API (FastAPI, MongoDB) that cut page load by 60% for 10K+ item inventories
- Integrated OAuth 2.0 SSO for enterprise accounts; built admin debugging dashboards that cut troubleshooting time for support tickets
Software Engineer
- Built Awaaz-e-Sehat on Flask and AWS Lambda (5s → 500ms API latency) with serverless ingest over 50K+ clinical recordings
- Productionized Whisper + GPT-4 transcription into structured clinical notes (95% accuracy, 1K+ recordings/day)
Publications
Uncovering the Hidden Data Costs of Mobile YouTube Video Ads
ACM Web Conference 2024 (WWW '24) • Singapore • May 2024 · doi:10.1145/3589334.3645496
First independent empirical study of mobile YouTube ad data costs from the user perspective. We streamed 17,600 main videos and 46,600+ ads (~8,225 hours) across 8 countries and showed latent buffer wastage — e.g. users still pay for 80–100% of a skippable ad's data ~31–53% of the time after skipping, mid-roll ads force re-download of ~71s of main-video on average, and excess losses average ~6.7% of a 2GB plan. Public dataset and toolchain released.
Get In Touch
I'm seeking software engineering / research internship opportunities for Spring / Summer 2027 — especially backend, infrastructure, distributed systems, and research engineering roles. Feel free to reach out.
Email: harrisah@buffalo.edu
Location: Buffalo, New York
GitHub: github.com/harris-ahmad
LinkedIn: linkedin.com/in/harris-ahmad1
ORCID: orcid.org/0009-0008-5402-8398