sharavana kumar

I’m a backend engineer focused on building cloud-native security systems—especially deception, threat intelligence, and resilient event-driven services. I work primarily with Go and Python to turn security research ideas into robust, test-backed infrastructure.

In my flagship project, I built two wire-protocol database deception proxies from scratch in Go, supporting PostgreSQL (pgproto3) and MySQL handshake v10. I also implemented an 8-step SQL normalization pipeline with FNV-64a fingerprinting to power Kafka/Redpanda partition keying, 10-second deduplication windows, and response-shaper rule lookups across both proxies, with 105 proxy tests passing.

I designed a 7-agent, event-driven microservice system (Go + Python) that communicates exclusively through Redpanda topics. The agents cover MITRE ATT&CK rule matching, DBSCAN session profiling, adaptive deception via FastAPI, and autonomous scaling—plus an MITRE intelligence agent with 10 YAML rules, 18 techniques, and an HMM Viterbi decoder scoring workflow (305 tests passing across 14 containers).

I’m also comfortable with production-minded operations and scaling: I implemented a Go scaling agent using a weighted confidence/depth/entropy/auth formula, EWMA smoothing, Welford Z-score bot suppression, hysteresis, and KEDA-driven autoscaling via the External Metrics API. Alongside this, I’ve done threat-hunting and AI-agent intensive work, mapping adversary behavior to MITRE ATT&CK and building structured, multi-agent LLM workflows using the A2A framework.