Associate Team Lead
- Lead a 9-engineer team owning all AI infrastructure and threat-intelligence systems end to end (architecture, execution, production), directly driving 90% of all downstream threat detections and automated client alerting.
- Architected and operate production Elasticsearch infrastructure indexing 170B+ records across 84TB of intelligence data, with custom analyzers, ILM policies and query optimization for sub-second latency on deep investigations.
- Engineered a high-throughput AI analysis pipeline processing 100K+ cases/day using local LLM inference (SGLang/Ollama), containerized with Docker and orchestrated through Redis/Celery task queues. Eliminated external API costs, saving $50K–$100K/year, with sub-100ms inference latency per case.
- Designed an end-to-end GitHub brand-abuse and data-leak detection pipeline for an enterprise client: keyword-driven preprocessing, context assembly (screenshots to S3, code to Qdrant embeddings), a multi-model inference stack (Qwen 7B for vision, Mistral 24B verification, Mistral 7B judge), Kafka topic partitioning, Redis queuing, PostgreSQL state and Kubernetes auto-scaling, reaching over 95% precision on newly leaked repositories.
- Optimized filtering across 100B+ combo records and 60B+ stealer records, cutting query latency by more than 70% with custom Elasticsearch templates and distributed pre-filtering.
- Re-engineered distributed Telegram ingestion and scraping workflows for 20x throughput (100–200 to 2K–5K files/day) through Kafka partitioning and worker-pool optimization.
- Built resilient multi-source monitoring agents for exposed infrastructure, S3 bucket leaks, .onion sites and GitHub repositories, automating dark web threat-surface discovery and client notifications.
