Most of my work.
Twelve systems, research, open-source, and applied AI projects. Select any row for the full engineering story.
Agentic Coding Reliability · NSF Cybersecurity-Funded
An NSF-funded study of whether software written by LLM coding agents is actually safe to ship. I rebuilt a classic UNIX reliability fuzzer into faster research tooling, tested a 12-program corpus of utilities including grep, make, and ptx, and root-caused real memory-safety and denial-of-service bugs.
- 12 programs · grep, make, ptx + more
- 4× faster evaluation tooling
- 10+ source defects investigated
- uutils ptx · 29× faster · 19% less peak memory
CS 220 Learning Infrastructure
I led the production systems behind CS 220 for 900+ students: scaled an AI tutor to 20K+ requests on peak days at 5× lower LLM cost, closed a root-level autograder RCE, and built observability that cut incident response time by 80% with zero-downtime releases.
- 900+ students · 20K+ peak requests/day
- 5× lower LLM cost
- 80% faster incident response
- Root-level RCE closed
Awaaz-e-Sehat
A React Native mobile app deployed in hospitals for seven months, using Whisper fine-tuned with LoRA on local Urdu speech to create structured records and surface dangerous maternal-health conditions.
- React Native · 500+ records
- 96% accuracy
- 60% less documentation time
- 7%→40% detection · 300+ alerts
Textish
A Python library that serves any Textual app over SSH through a custom terminal driver, with isolated app instances coordinated on one asynchronous event loop.
- 96× lower memory
- 4.3× faster startup
- 4 PyPI releases
- Load-hardened SSH service
NVIDIA Sirius DB
Added string aggregation to NVIDIA Sirius DB, unblocking a full industry benchmark suite and correcting overflow and empty-value calculations in GPU execution.
- 1K+ star project
- Full benchmark suite unblocked
- Overflow + empty-value fixes
- 2 merged PRs
Gemini CLI History Search
I added two production features to Gemini CLI: cached, debounced Ctrl+R prompt history search and an expandable multiline prompt renderer, merged upstream across two pull requests.
- 100K+ star project
- 2 merged PRs
- Cached incremental search
- Multiline prompt renderer
Designing Around Stigma
A WhatsApp LLM chatbot for menstrual-health education, grounded in expert-curated content to minimize hallucinations and evaluated in a low-resource setting.
- 403 deployment messages
- 2-week deployment
- AWS EC2 + DynamoDB
- ACM CHI 2026
Evaluating Linux Schedulers
An empirical comparison of Linux EEVDF and CFS across latency-sensitive and CPU-bound workloads, combining application-level measurements with eBPF traces under controlled contention.
- EEVDF · 4–6× lower p99.9 wakeup latency
- EEVDF · up to 65% higher Redis GET throughput
- CFS · 13–59% lower HTTP latency
- CFS · 10–15% higher 7-Zip throughput
VIDER Video Reconstruction
A model-agnostic streaming framework that sends sparse keyframes over long-haul links and reconstructs intermediate video frames near the viewer.
- 97.69% avg. bandwidth savings
- 0.83 avg. ViSiL similarity
- 28-video evaluation
- Beat H.264 CRF 50 on savings + similarity
Food Insecurity Forecasting
A replication and Pakistan-focused extension of a Science Advances food-crisis forecasting study, comparing Random Forest, panel regression, ordinal regression, and LLM-agent approaches over Urdu news and weather data.
- 94% IPC prediction accuracy
- 90× faster RF training
- 162,653 Urdu articles
- RF · OLS/Lasso · LogisticAT · LLMs
BookSwitch
Arc is my favorite browser, but it doesn't run on Linux—and what I missed most was jumping to a pinned site with a single keystroke. BookSwitch rebuilds that shortcut for Chrome and Firefox: Alt+1 through Alt+9 open the first nine bookmarks in your bar. Published on both stores.
- Published on both stores
- One build, two browsers
- Alt+1–9 bookmark jumps
- No new list to maintain
Slack AI Course Bot
A context-aware Slack assistant that grounds Gemini responses in CS 382 assignment materials and supports course help, anonymous posts, and channel activity tools.
- 4 Slack commands
- Assignment-grounded Gemini
- Thread-aware responses
- Rate-limited async service