Yazeed Shamsudeen

My career started in computer vision at TechAtPlay, where I built real-time sports analytics systems using YOLOv5, OpenCV, and PyTorch. I designed CI/CD pipelines with GitHub Actions, built FastAPI backends for tracking and trajectory estimation, and developed custom YOLOv5 training workflows for cricket shot-direction analysis — including pose detection with Mediapipe and YOLOv8.

From there, I joined SOTI Inc. as a full-stack Software Developer, where my most significant achievement was architecting a high-performance in-memory search engine from scratch to replace Elasticsearch. The system reduced query latency from 800ms to under 50ms across 1M+ device deployments. I built a custom indexing layer using compressed bit-sets and B-tree structures for sub-millisecond Boolean filtering, and designed a watermark-based incremental sync pipeline to keep SQL Server and the in-memory layer in real-time consistency. I also implemented a fault-tolerant HA architecture with master-failover replication, mutual TLS, and atomic disk persistence — achieving zero-downtime deployments for enterprise MDM clients.

Beyond my day job, I built an AI coding agent (similar to Bolt/Replit) using Next.js, Claude API, Convex, and Inngest — it generates code from prompts, scrapes URLs for live context, previews in real time, and supports GitHub import/export. It’s live in production with Stripe payments, Clerk auth, and Sentry monitoring.

My technical interests lie in distributed systems, developer tooling, AI-powered applications, and cloud infrastructure. I’m comfortable across Python, C#, TypeScript, and React, and have hands-on experience with AWS (Lambda, Aurora, SQS, DynamoDB), Kafka, Redis, Kubernetes, and Terraform. I’m particularly drawn to companies building foundational infrastructure or applying AI to solve real engineering problems.​​​​​​​​​​​​​​​​