Backend Developer & Team Lead
Go - TCP/IP - Symfony - Doctrine ORM - Laravel - PHP-Swoole - HAProxy - Nginx - MariaDB - Redis - Meilisearch - Kafka - Python - FastAPI - Qdrant - RAG - rembg - Node.js/TypeScript - AWS EC2/S3/ECR - Prometheus - Grafana - Docker Swarm - Kubernetes
- Led 3-person backend team delivering REST and GraphQL (Laravel Lighthouse) APIs across 4 interconnected mobile apps (grocery, food delivery, merchant management, recipe reels), coordinating domain boundaries across Symfony and Laravel services.
- Built and owned a high-performance Go TCP server processing 5,000,000+ delivery-driver location updates per day, plus thousands of other real-time events (stock availability, order notifications), forming the core of a dispatch pipeline through a Kafka cluster (set up and operated end-to-end: broker config, topic/partitioning design, consumer groups) and a matching algorithm to the driver's app.
- Prevented race conditions in concurrent order acceptance, ensuring exactly one driver claims each order - matched to the nearest available driver using real-time location data.
- Reduced product search latency to sub-50ms across 70,000+ SKUs by designing a Redis + Meilisearch search layer with query caching, replacing slow SQL full-text scans.
- Increased concurrent user capacity 6x by replacing synchronous PHP-FPM with PHP-Swoole coroutines, enabling peak load without horizontal scaling.
- Improved product discovery conversion by 30% by designing a personalized recommendation algorithm optimizing display rank across 70,000+ products using purchase history and catalog signals.
- Surfaced ranking signals across 200+ store catalogs by engineering a top-selling product algorithm aggregating sales velocity and inventory data for marketplace dashboards.
- Applied DDD to decompose a monolithic Q-commerce codebase into bounded contexts (catalog, ordering, fulfillment), reducing cross-team merge conflicts and enabling independent deployments.
- Built and maintained Python FastAPI async REST APIs over ~3-4 years, including a Qdrant vector database-backed recipe recommendation service powering the recipe reels app, surfacing similar recipes based on users' liked reels.
- Extended the recipe recommendation service into a RAG pipeline, retrieving similar recipes via Qdrant vector search and generating personalized descriptions with an LLM.
- Automated product image background removal using rembg, integrated into the product-catalog pipeline to standardize listing images at scale.
- Deployed and operated backend services on Kubernetes, managing rolling deployments, scaling, and incident troubleshooting alongside Docker Swarm-based workloads.
- Built an internal Node.js/TypeScript admin/dashboard tool giving operations teams visibility and control over backend data and workflows.
