Experience
2care.ai — Backend & AI Engineer
Sept 2026 – Oct 2026
TypeScript · Next.js · AWS (Lambda, Step Functions, Aurora, Amplify)
- Led the insurance verification feature end to end, pairing voice AI agents for conversational data collection with a deterministic rules engine that keeps LLM non-determinism out of every verification decision.
- Built the verification workflow on serverless AWS (Lambda, Step Functions, Aurora) and validated the rules engine against 46K+ test cases, achieving 100% verification accuracy.
- Improved the voice agent’s turn-taking and interruption handling, fixing mid-sentence cut-offs so conversations sound natural and human-like.
Proqio — Software Engineer (Backend / Distributed Systems)
Dec 2023 – Aug 2026 · Spain (Remote)
Node.js · TypeScript · Kafka · PostgreSQL/TimescaleDB · Kubernetes
- Architected a real-time IoT structural-health-monitoring platform (tunnels, bridges, dams) as 14+ event-driven microservices (Kafka across 12 services, PostgreSQL/PostGIS) with end-to-end correlation-ID tracing.
- Built the multi-protocol data-ingestion gateway integrating 3 external monitoring vendors — RST Instruments, Leica GeoMoS, Sigicom — over REST polling, MQTT, and FTP/SFTP, normalizing 6 geotechnical sensor types into a unified time-series schema.
- Engineered a Postgres-backed distributed polling scheduler dispatching up to 500 sensor-poll jobs/cycle at 1-minute granularity with retry/back-off, offloading file parsing to a worker-thread pool to keep ingestion non-blocking under load.
- Owned time-series at scale: migrated PostgreSQL → TimescaleDB and added DuckDB for analytics over 200M+ readings, cutting dashboard query latency by ~60%.
Infosys — Full Stack Developer
June 2022 – Dec 2023 · Bangalore
Java · Spring Boot · MySQL
- Optimized RDBMS performance on a core Airbus tool by refactoring legacy stored procedures and redesigning schemas, cutting report-generation time 14s → 3s.
- Re-architected data fetching for a high-accountability internal tool, slashing search latency 3.5s → 100ms; drove Git-flow and system-level scalability best practices.
Internships — HoloWorld · SalesBook
July 2021 – June 2022
- HoloWorld — Node.js/MongoDB backend. SalesBook — backend for 100K+ users.
Selected Projects
Node.js · WebSockets · vLLM · Python
A fully-streaming speech-to-speech assistant with overlapping STT → LLM → TTS stages over a single WebSocket, reaching ~165 ms time-to-first-audio ~10× faster than V1 on one 12 GB GPU.
Evolved it monolith → RabbitMQ queue → streaming: energy gate (no VAD), streaming faster-whisper STT, vLLM (continuous batching), Kokoro TTS, barge-in, and conversation memory.
RnC (Research & Coding) Assistant ↗ GitHub
Python · LLM Agents · ReAct · MCP
An autonomous research-and-coding agent built from scratch (no framework) on a pluggable skill registry — web research, file I/O, and sandboxed code execution — driven by a ReAct loop with multi-step planning and a prompt eval harness.
Architected as an extensible multi-agent system — handoffs to specialized sub-agents, persistent memory, and MCP tool connectivity — model-portable across OpenRouter, local vLLM, and Ollama.
codecrafters-redis-rust ↗ GitHub
Rust · Systems programming
A Redis server implemented from scratch in Rust — RESP protocol parsing, a key/value store with expiry, and concurrent client handling.
Skills
Languages: TypeScript, Python, Java, SQL, C/C++
Backend: Node.js, Next.js, Express, REST, WebSockets, Kafka (event-driven), RabbitMQ, gRPC, MQTT
Data: PostgreSQL, Amazon Aurora, PostGIS, TimescaleDB, DuckDB, Redis, MinIO/S3 (time-series at scale)
AI / LLM: LLM & voice AI agents, tool-calling, multi-agent orchestration, RAG, LangChain, LangGraph, vLLM
Cloud / DevOps: AWS (Lambda, Step Functions), Docker, Kubernetes, Helm, ArgoCD, Istio, Keycloak, OpenTelemetry