Chinton Dutta
AI Full-Stack & DevOps Engineer
About
I'm an AI full-stack and DevOps engineer focused on building production-ready AI systems and cloud-native backend platforms. I work across Python, TypeScript, and Next.js, designing multi-tenant SaaS architectures, RAG and LLM pipelines, and Kubernetes-based infrastructure with ArgoCD and Helm.
As a Founding Engineer at FREMN, I lead backend architecture and AI infrastructure for a multi-tenant healthcare SaaS platform, including fine-tuning voice AI models for production use. I'm currently open to AI full-stack and DevOps engineering opportunities.
Experience
Nov 2025 – Present
Kolkata, India
Built and deployed a multi-tenant AI SaaS for dental practice management, including WhatsApp-native booking and automated missed-call recovery. Architected Kubernetes infrastructure with ArgoCD and Helm for zero-downtime releases, and fine-tuned Chatterbox TTS and Whisper STT using LoRA to improve voice agent accuracy for Indian accents and dental terminology.
Aug 2025 – Nov 2025
Kolkata, India
Owned backend API development for client-facing products, designing relational schema and RESTful endpoints to support new features end-to-end. Architected multi-tenant backend infrastructure for a hospitality management system using tenant-ID-based data isolation, and maintained Docker-based CI/CD pipelines while leading code reviews to keep production environments stable.
Projects
Engineered a multi-tenant TTS SaaS where users clone voices from audio uploads, storing files in Cloudflare R2 and mounting them directly into Modal A10G containers to keep the app server fully stateless. Deployed Chatterbox TTS as a FastAPI ASGI app on Modal GPU, supporting 10 concurrent requests with persistent model loading that cut cold-start latency by roughly 80%.
Solvia
LiveAI-Powered Customer Support Platform
Built a multi-tenant AI support platform with a per-organization RAG knowledge base and automatic AI-to-human escalation, cutting first-response time to under 3 seconds across all incoming chats. Designed a real-time operator dashboard on Convex's reactive backend, synchronizing live conversation updates across the widget and dashboard without building a custom WebSocket layer.
Architected a multi-agent GPT-4.1 pipeline that decomposes natural language prompts into fully deployed, multi-file Next.js applications, using Inngest to durably orchestrate generation steps without losing state on failure. Integrated E2B cloud sandboxes to isolate and hot-reload generated code in real time, streaming a fully working live preview URL into the chat UI in under 90 seconds.
Skills
Languages
AI / ML
Frontend
Backend
Databases
DevOps
Education
Sep 2023 – Aug 2027
Kolkata, India
Techno India University
BCA(H) — Bachelor's in Computer Application (Honours)
Contact