Headline

AI Engineer | Voice AI | Real-Time AI Agents | Generative AI | LLMs | Fine-Tuning | LiveKit | Python

About

I’m an AI Engineer focused on designing and building production-grade AI systems, with a strong focus on Voice AI and Real-Time AI Agents. My work spans the full AI engineering lifecycle — from system architecture and LLM orchestration to backend services, real-time communication, cloud deployment, and production monitoring. I work extensively with Voice AI, LLMs, fine-tuning, tool calling, RAG, and agentic workflows. I build real-time conversational systems using technologies such as LiveKit, Python, FastAPI, LangChain, and LangGraph. My key areas of work include: • Voice AI & Real-Time AI Agents • LLM applications, fine-tuning & evaluation • Agentic systems, tool calling & MCP • System Design & AI Architecture • Backend APIs & distributed AI workflows • Real-time and low-latency systems • Cloud deployment, Docker & production infrastructure • Observability, reliability & scalability I enjoy solving the engineering challenges that come after the AI prototype — designing the right architecture, reducing latency, managing context and state, integrating external systems, scaling services, and making AI applications reliable in production. I’m particularly interested in building intelligent, real-time AI systems that combine models, voice, tools, backend infrastructure, and cloud services into complete products.

Location

India

AI stack

claude code

hugging face

langchain

langgraph

livekit

vLLM

Work experience

AI/ML Engineer at Trugen AI, Feb 2026 - Present · 9 mos, Remote, - Agent Harness Ownership: Owned the architecture and hands-on development of a production AI-agent harness platform handling long-running tasks, background workers, context, tools, state, checkpoints, limits and traces. - Led Model Finetuning for Production: Led team to prepare labelled data and held-out evaluation workflows, then fine-tuned and deployed Qwen3.5-0.8B, achieving 97.9% accuracy and 0.975 Macro F1. - Observability, Latency, and Cost: Built distributed tracing across agent turns, tool calls, and background workers using OpenTelemetry and Langfuse; monitored request-level latency, token burn, and error rates to debug edge-case failures and control inference spend. - Real-Time Voice & Video Agents: Architected and deployed interactive multi-participant meeting bots using LiveKit, Recall.ai, Python, and streaming speech (STT/TTS); orchestrated low-latency RPC and tool calling across bidirectional voice, screen-share, and video streams. - Streaming Infrastructure: Re-architected event ingestion into an auto-reconnecting, call-scoped WebSocket pipeline, eliminating a 10-second event blind spot during agent handoffs and safeguarding session state against partial network drops. - Reusable AI Infrastructure: Built authenticated, OpenAI-compatible APIs with Server-Sent Events and multi-provider routing; developed a Go and Gin service to provision, monitor, reconcile, and manage Dockerized agent runtimes.

ML Engineer - GenAI at FutureSmart AI, Oct 2024 - Feb 2026 · 1 yr 5 mos, - System and Multi-Agent Delivery: Led end-to-end delivery of Generative AI, custom-agent, and NLP systems for global clients, translating business requirements into architectures, prototypes, and production deployments while remaining hands-on in development. - Agent, RAG, and Integration Architecture: Designed and built multi-agent workflows, modular pipelines, third-party integrations, and retrieval systems combining SQL and vector databases for real-time, context-aware client workflows. - Production and Cloud Delivery: Deployed and scaled Dockerized applications on AWS and GCP, improving reproducibility across development and production environments. - Team Mentorship: Mentored interns and junior engineers on software engineering, system design, and prompt engineering; recognized multiple times as Employee of the Month.

Data Science Intern at FutureSmart AI, Mar 2024 - Oct 2024 · 8 mos

Education

Bachelor's degree, University of Calcutta