TruGen AI
Real-time voice classifier
Fine-tuned a hybrid GatedDeltaNet / full-attention transformer (Qwen3.5-0.8B) as a binary classifier for real-time voice agents.
97.9% acc · 0.975 Macro F1
Fine-TuningQwen3.5LiveKit
Available for AI / ML roles & collaborations
> ai_ml_engineer
Rounak Show
Building real-time voice & video AI agents, fine-tuned LLMs, and multi-agent systems that drive business impact.
2+ yrs
shipping production AI
97.9%
voice classifier accuracy
AWS
Certified Cloud Practitioner
selected_work
Things I've built
Production AI systems across real-time voice, agents, and retrieval — from fine-tuned models to full pipelines.
TruGen AI
Meeting-bot pipeline
Lets AI agents join Zoom, Google Meet and Microsoft Teams live, with WebSocket-based transcript and event streaming.
WebSocketReal-TimeLiveKit
TruGen AI
Vision + MCP agent
Screen-share agent that interprets shared screens live, with MCP-based tool calling for reliable session and tool routing.
VisionMCPAgents
FutureSmart AI
Text-to-SQL retrieval
Turns natural-language questions into validated SQL over client warehouses, backed by SQL + Vector DB context retrieval.
NL2SQLRAGLangGraph
FutureSmart AI
Multi-agent support
RAG-powered customer-support agents resolving tickets with cited, domain-specific answers for global clients.
RAGMulti-AgentVector DB
FutureSmart AI
Résumé ranking & parsing
Parses and scores candidates against a role spec to speed up screening, built as a modular, production-ready service.
NLPPythonFastAPI
about_me
AI / ML Engineer
I specialize in real-time conversational AI. At TruGen AI, I fine-tune and deploy custom models into production voice pipelines, and build meeting-bot integrations that let AI agents join live, multi-participant meetings.
My background spans Generative AI application development, retrieval-augmented generation, LLM fine-tuning, and scalable cloud deployment on AWS and GCP. I'm passionate about applied AI, real-time systems, and writing technical content that drives innovation.
Howrah, West Bengal, IndiaEmployee of the Month, multiple times
tech_arsenal
Skills & tools
Languages & Frameworks
PythonFastAPILangChainLangGraphLiveKit Agents SDK
Generative AI
LLMsFine-Tuning (LoRA / PEFT)RAGMulti-Agent SystemsKnowledge GraphsPrompt Engineering
Real-Time Voice AI
LiveKitWebSocket StreamingSTT / TTS OrchestrationMeeting-Bot Integrations
MLOps & Cloud
AWSGCPDockerGitHub ActionsRunPod (GPU)
Data & Storage
SQLNoSQLVector Databases
career
Experience
AI/ML Engineer
TruGen AI
CurrentFeb 2026 – Present
- Fine-tuned a hybrid GatedDeltaNet/full-attention transformer (Qwen3.5-0.8B) as a binary classifier for real-time AI agents — 97.9% accuracy and 0.975 Macro F1 on held-out test data.
- Built a meeting-bot pipeline enabling AI agents to join external multi-participant meetings (Zoom, Google Meet, Microsoft Teams) in real time, with WebSocket-based transcript and event streaming.
- Shipped a vision-capable screen-share agent that interprets shared screen content live, and integrated MCP-based tool calling to improve session handling and tool routing reliability.
PythonLiveKitFine-TuningWebSocketMCPMulti-Agent
AI Engineer
FutureSmart AI
March 2024 – Jan 2026
- Led end-to-end AI projects for global clients, delivering Generative AI solutions, custom agents, and NLP-driven applications — rapidly prototyping and deploying production-ready MVPs.
- Designed multi-agent workflows, modular pipelines, and third-party integrations (Odoo, databases), with retrieval systems using SQL + Vector DBs for real-time, context-aware responses.
- Deployed and scaled applications on AWS and GCP with Docker, mentored interns and junior engineers, and was recognized multiple times as Employee of the Month.
PythonFastAPILangChainLangGraphRAGAWSGCPDocker
writing
Technical blogs
Sharing what I learn about AI, LLMs, and agentic systems.
