Dev Sharma
Backend & AI Engineer
*********.***@*****.*** +91-960******* linkedin.com/in/devtalksagi SUMMARY
Backend & AI Engineer with 4+ years building production Python systems and LLM applications. Designed a multi-agent document-extraction pipeline (LangChain, Celery) serving 10+ enterprise healthcare clients, and built RAG and MCP-based tooling in production. First engineering hire at a healthtech startup, where I built the backend from zero. Comfortable owning the full path from API design to deployment on GCP and AWS. PROFESSIONAL EXPERIENCE
Docstribe, Backend & AI Engineer 10/2025 – Present Gurugram, India
- Re-architected a single-API OCR flow into a multi-agent LangChain pipeline with Celery workers, combining multiple AI models and async processing to improve prescription extraction accuracy and speed across 10+ client environments.
- Built RAG pipelines over internal healthcare data to power AI-driven workflows for enterprise clients.
- Built MCP servers exposing internal tools and data to LLM agents, enabling agent-driven workflows across the platform.
- Took ownership of AI, infrastructure and security for a platform serving 10+ enterprise clients: CI/CD with Terraform on GCP, Cloud SQL moved into a private VPC, automated client deployments, and Metabase on Cloud Run against a read replica for self-serve analytics. Mizzle Enterprise, Python Developer 05/2025 – 09/2025 Gurugram, India
- Primary backend developer for a cloud platform, owning payments, FastAPI services and third-party integrations (including Jira for task and team management), and worked on AI/LLM integrations.
- Integrated third-party GPU APIs for provisioning, configuration and log monitoring; when the provider had no log API, negotiated access to an SSE-based stream and integrated it, avoiding a lengthy custom build. Babynama, Backend Engineer (First Engineering Hire) 05/2022 – 03/2025 Gurugram, India
- First engineering hire; built the backend from scratch in Python (Django, FastAPI) and grew it into the company's core platform over 3 years.
- Digitized the full pediatric consultation workflow (appointment booking, doctor slot management, payments), replacing paper and Excel-based processes.
- Designed REST APIs and schemas on PostgreSQL, MongoDB and Redis for async notification pipelines, payment gateways and real-time consultation tracking.
- Containerized services with Docker and deployed on AWS (EC2, S3, Lambda) with CI/CD, owning infrastructure from development to production.
PROJECTS
Ram Mandir Knowledge Assistant, Multilingual RAG (personal project)
- Question-answering system over Hindi and English sources using hybrid retrieval (BM25 + embeddings) and reranking, returning answers with citations.
- Evaluation set to compare chunking and retrieval strategies and measure answer quality; served via FastAPI and deployed on Cloud Run.
SKILLS
Languages: Python, SQL
AI / LLM: RAG, LangChain, LangGraph, MCP, multi-agent systems, vector databases, prompt engineering, LLM evaluation
Backend: FastAPI, Django, Celery, REST APIs, async Python Databases: PostgreSQL, MongoDB, Redis
Cloud & Infra: GCP (Cloud Run, Cloud SQL, VPC), AWS (EC2, S3, Lambda), Docker, Terraform, CI/CD EDUCATION
Chandigarh University, B.Tech 07/2018 – 07/2022 Chandigarh, India