Resume
Download PDF ↓Manas Rai — Software Engineer · GenAI Engineer
Bengaluru, India · rai.manas12@gmail.com · GitHub · LinkedIn
Software engineer with ~5 years building production backend systems, the last ~2 focused on Generative AI — LLM applications, RAG pipelines, and multi-agent systems — on a foundation of async Python microservices, distributed data pipelines, and semantic search. Owns systems end-to-end: architecture, backend services, cloud infrastructure, and production launch, including a multi-tenant healthcare RAG platform load-tested for 2,000+ concurrent sessions at sub-second latency.
Skills
- Backend & languages: Python, Go, FastAPI, Flask, Django, REST APIs, Microservices, WebSockets, Async programming, SQLAlchemy, Alembic, System design
- Generative AI & LLMs: LLMs, RAG, AI agents, Multi-agent orchestration, Prompt engineering, LangChain, LangGraph, LangSmith, OpenAI API, Anthropic Claude API, Model Context Protocol (MCP), Google ADK, A2A Protocol, Embeddings, Vector databases, LLM evaluation (RAGAS), LLM observability & cost monitoring
- Data science & ML: pandas, numpy, scikit-learn, EDA, Feature engineering, Hugging Face
- Frontend: React, JavaScript, HTML5
- Databases & vector stores: PostgreSQL, MySQL, MongoDB, Neo4j, Snowflake, Aurora, Redis, Pinecone, pgvector, FAISS, MongoDB Atlas Vector Search
- Cloud & DevOps: AWS (EC2, RDS, ECS, S3, SQS, Lambda, Bedrock, Cognito), Azure OpenAI, Docker, Kubernetes, Jenkins, CI/CD, Datadog, MLOps
- Tools & practices: Claude Code, VS Code, Jira, GitHub, Agile, Scrum
Experience
- Launched a healthcare chatbot from zero to production by architecting a multi-tenant RAG pipeline on Azure with tenant-isolated vector stores, secure authentication, and data partitioning — directly enabling the product's first paying customers.
- Engineered an LLM-powered clinical simulation platform using prompt engineering and RAG to replicate real patient interactions for physician training; load-tested for 2,000 concurrent sessions at sub-second response latency.
- Built a multi-tenant VS Code agent platform with a meta "maker" agent that generates governed agents, skills, hooks, and prompts — interviewing the developer, planning, and checking new instructions against the existing set before implementing.
- Reduced manual developer workflow time by ~60% by automating multi-step SDLC tasks (code-review scaffolding, test generation, documentation) through agentic orchestration patterns.
- Shipped a multi-tenant lead-generation data pipeline with SQLAlchemy and AWS Cognito, enforcing per-account data isolation across 5+ client accounts.
- Built an automated NLP-powered web crawler that extracts and structures institutional data, powering a qualification-matching recommendation engine — cutting 15+ hours/week of manual research.
- Promoted 3 times in 3 years (Trainee → Developer → Senior Developer → Solution Leader), the fastest progression in the engineering org at the time.
- Cut data retrieval time by 75% (4x faster) by implementing semantic search with Neo4j graph traversal and MongoDB vector indexing across 500K+ embeddings and data points.
- Improved real-time audio transmission stability by 35% by engineering bidirectional WebSocket streaming with adaptive buffering and error recovery for NLP voice applications.
- Designed and deployed async Python microservices for non-blocking concurrent request handling, increasing API throughput and resolving request-timeout failures seen under production load.
- Built a FastAPI persistence service with JWT authentication serving as the data backbone for multiple product interfaces, with zero-downtime deployments and CI/CD integration.
Selected projects
-
DevFlow Kit
Python · LangGraph · GitHub Actions · Claude Code · Jira
Multi-agent SDLC automation that turns Jira tickets into production PRs with zero added infrastructure. Refinement, implementation, and Jira-sync agents decompose complex tickets into parallel subtasks and cut the ticket-to-PR cycle from days to hours.
-
RegLens
Python · LangGraph · Google ADK · RAGAS · pgvector
Multi-agent regulatory compliance automation. Feed it a regulatory PDF and your control matrix — a compliance research agent extracts every obligation, a gap analyzer checks each against your policies via RAG and scores the risk, and a report generator produces an audit report with a human-in-the-loop approval gate. Includes a drift-detection evaluation harness.
-
CostTracker
Python · ClickHouse · PostgreSQL
Open-source, self-hosted LLM cost tracking SDK. A drop-in instrumentation layer wraps OpenAI, Anthropic, Groq, and Bedrock clients to record usage straight to ClickHouse or PostgreSQL — real-time token metering, per-request cost attribution, and a bundled analytics dashboard.
-
Cloud Waste Hunter
Python · FastAPI · scikit-learn · Next.js
Cloud-agnostic resource monitor with ML-powered waste detection. Flags idle instances, unattached volumes, and stale snapshots across providers in a unified cost-optimization dashboard — then eliminates them safely with dry-run previews, human-in-the-loop approval, and 7-day rollback.
Certifications
- Building with the Claude API — Anthropic Academy
- Introduction to Model Context Protocol (MCP) — Anthropic Academy
- Claude Code in Action — Anthropic Academy
Awards
- Outstanding Performer Award — Tricon Infotech (Dec 2025), for delivering 2 end-to-end AI products in a single year.
- Spot Award — Tricon Infotech, for shipping the healthcare chatbot from architecture to production launch.