-0284C7?style=flat-square&logo=google-scholar&logoColor=white){kind=icon}
I'm a Computer Science and Statistics student at the University of Toronto (Honours B.Sc., Co-op) and a 3x hackathon winner. I enjoy building across the stack, from low-level systems in C++ to production Node/FastAPI applications and applied AI/ML tooling. I care about reliability, performance, and clean developer experience, whether that means shaving lock contention out of an embedded key-value store or cutting API failure rates in a production service.
🎯 Open To: Software Engineering Internships · Software Engineer · Software Developer · Full-Stack & Backend Engineering · AI/ML Engineering
Languages
Frameworks & Databases
Cloud, DevOps & Tooling
Robotics
🗃️ Embedded Key-Value Store — C++ / CMake / Multithreading / Linux
An embedded C++ database engineered for high-throughput concurrent access via sharding and lock-free reads.
| Stack | C++, CMake, Multithreading, Linux |
| Scale | 16-way sharded locking, 5x read throughput (8M → 40M ops/sec) |
| Performance | 45%+ disk usage reduction via compaction |
| Reliability | Checksummed write-ahead log for crash recovery |
| Repository | github.com/ZayaanB |
- Built an embedded database with 16-way sharded locking, boosting read throughput 5x (8M to 40M ops/sec)
- Implemented reader-writer locks so reads proceed during disk writes, preventing race conditions across 16 threads
- Created a file compaction routine to safely clean up keys, reducing disk usage by 45%+ without pausing the app
- Implemented crash recovery by persisting every update to a checksummed write-ahead log before applying changes
🔗 Context Sync Extension — TypeScript / VS Code Extension API
A VS Code extension automating AI chat context continuity across developer environments — 750+ downloads.
| Stack | TypeScript, VS Code Extension API |
| Scale | 750+ downloads |
| Performance | 25%+ token-use reduction, 30%+ summary compaction |
| Repository | github.com/ZayaanB |
- Developed a VS Code extension with 750+ downloads that automates AI chat context sharing across environments
- Reduced token use by 25%+ by modelling chats as a weighted graph and selecting context via shortest-path search
- Compacted chat summaries by 30%+ by designing a Markdown schema to pack more context into fewer tokens
🏥 Clinical AI Assistant — Python / SQL / OpenCV — Top 10 @ GenAI Genesis
An AI-powered clinical assistant automating patient intake with real-time computer-vision monitoring.
| Stack | Python, SQL, OpenCV |
| Scale | Full intake-to-monitoring pipeline |
| Performance | 98%+ real-time fall detection accuracy |
| Recognition | Top 10 Projects @ GenAI Genesis |
| Repository | github.com/ZayaanB |
- Deployed an AI-powered clinical assistant automating patient check-ins/intake with structured data schemas
- Engineered a live monitoring dashboard using computer vision for real-time fall detection
- Built a spatial visualization algorithm generating 3D labelled hospital models by extruding floor plans
Software Developer Intern (Full Stack) · FreshBooks
Sept 2026 – Dec 2026 · Toronto, ON
- Accelerated bank connections page loads by 86% (700ms to 100ms) by removing an inactive-bank API fetch.
- Reduced invoice email latency by 1–3s by decoupling PDF generation across services with GCS-backed event delivery.
- Improved backend execution speed by 13.6x across 20 benchmark calls by caching metadata and invoice lookups.
- Enabled proactive bank reconnection warnings by tracking Yodlee consent expiry across webhooks, storage, and APIs.
- Unified Plaid and Yodlee bank connections across backend services and Ember for staged rollout to 100% of users.
Python FastAPI Ember.js GCP
Autonomy Software Engineer (Mapping & Planning) · University of Toronto Formula Racing – Driverless
July 2026 – Present · Toronto, ON
- Developed mapping and path-planning software in C++ and ROS 2 for Canada's first driverless FSAE car.
- Replaced FastSLAM with EKF-SLAM, reducing localization compute time by 30% while maintaining accuracy.
- Built ego-aligned track corridors from quintic splines using arc-length sampling across a configurable 15 m horizon.
- Expanded simulation data collection and monitoring by integrating 8 sensors, including IMU, wheel-speed, and GPS.
- Cut ~20 ms of planning latency by switching to LiDAR-only planning and removing the camera neural network.
Python C++ ROS2 LiDAR
Software Engineer Intern (Backend AI) · FlyRank AI
July 2026 – Sept 2026 · Toronto, ON
- Prevented duplicate charges under 50+ simultaneous requests by deduplicating payment events in the database.
- Built a centralized embeddable widget platform, routing form submissions to a dashboard via a one-line script tag.
- Secured a public endpoint with validation, rate limiting, and geolocation checks, cutting spam by 80%.
TypeScript Node.js Express.js SQLite Zod
Software Developer Intern (Full Stack) · KorraNet Creative
May 2026 – July 2026 · Remote
- Optimized Meta OAuth token handling and storage, reducing token refresh failures by 40%+ for AI integrations.
- Reduced system crashes and failed API calls by 85%+ by adding error handling and retries for failed requests.
Python FastAPI OAuth Google Cloud Platform
"Build systems that don't wake you up at 3am."