Sumit Devraye

Engaged in design, development, and deployment of Generative AI systems for enterprise applications across education, HR, and internal tooling use cases, focusing on LLM orchestration, document processing, and intelligent chat interfaces.

● Designed and deployed GenAI-powered tutoring and document tools, including a document merging system that preserves semantic flow across source files with topic-aligned restructuring.
● Developed a multi-agent AI tutoring assistant using RAG pipelines, LangGraph, and AutoGen, enabling contextualized learning responses and educational guidance.
● Engineered a system to analyze client-uploaded documents — including meeting transcripts and discussion notes — to infer and structure personalized learning needs for enterprise clients.

● Created a hierarchical mind map generation engine from unstructured documents using agentic orchestration and LLM-based chunk analysis to infer topic structure.

● Built evaluation modules combining embeddings and LLM feedback to automatically score outputs for completeness, hallucinations, and structure fidelity.
● Delivered an intelligent role recommendation and comparison platform using persona mapping and vector-based search on structured role data.
● Deployed and maintained LLM-backed services on AWS EKS and Azure App Service, managing CI/CD with Azure DevOps in a 5-member agile team.
● Leveraged frameworks such as LangChain, LangGraph, AutoGen, and pgvector to orchestrate scalable, production-ready GenAI solutions using GPT-4.1, Claude, and Falcon APIs.

● Automated ingestion and parsing of large-scale structured Excel inputs into PostgreSQL using Azure Event Grid and Blob Storage triggers.

● Demonstrated innovative GenAI systems to leadership teams, contributing to internal innovation showcases and adoption across business units.

Independent Prototypes & Explorations

● Podcast Generator from Documents: Built a prototype that transforms unstructured content into a podcast-style conversation between a host and expert using LLMs and Azure TTS. Explored persona modeling, prompt chaining, and multi-speaker synthesis.

Worked as a backend and cloud developer on a global billing platform serving international clients. Contributed to secure, scalable feature development while supporting production reliability and cloud automation across enterprise environments.

● Developed and maintained backend and frontend components for a revenue-critical billing system using Ruby on Rails, PostgreSQL, and Redis.
● Delivered end-to-end features — from design and development to testing and deployment — within an agile DevOps team.
● Automated Excel upload workflows using Falcon-based APIs, Azure Blob Storage, and Event Grid triggers, improving operational efficiency.
● Ensured production stability and performance through bug fixes, CI/CD pipeline improvements, and support across AWS infrastructure (EC2, S3).
● Collaborated with QA and infrastructure teams to release secure, stable builds across Citrix and Microsoft Exchange environments.
● Received consistent recognition from team leads for high-quality code, reliability, and long-term maintainability contributions.