- End-to-End Architecture: Developed the complete ecosystem from scratch, including cross-platform mobile apps (Flutter), web frontend, and a high-performance FastAPI backend integrating WebRTC (LiveKit) for live streaming.
- Resource Optimization: Orchestrated a production-grade ML infrastructure running 6 concurrent models (MiniLM, CLIP, NudeNet, Gradient Boosting) on a highly constrained 6 vCPU / 24GB RAM environment, keeping peak memory under 40% via ARQ async cron jobs.
- Multimodal Search & Discovery: Engineered a sub-50ms hybrid semantic search engine using FAISS IVFFlat for ANN retrieval and PostgreSQL pgvector for accurate re-ranking.
- Real-time Algorithms & Observability: Developed feed personalization utilizing Alternating Least Squares (ALS) and Thompson Sampling. Built a comprehensive observability stack (Grafana, ClickHouse, Loki) to monitor SSD IOPS and system bottlenecks.
- CI/CD & Platform Safety: Streamlined mobile deployments via Fastlane pipelines and integrated real-time automated moderation systems (NudeNet CNN, NLP filtering) to ensure continuous platform safety and compliance.