I'm an AI Engineer & Python Developer building multi-agent systems — LLM agents, RAG pipelines, and production-grade AI automation.

• 🏆 Open-source contributor to Cognizant AI Lab's Neuro SAN — 2 PRs merged to main
• 🤖 Building multi-agent networks with Neuro SAN, LangChain & Gemini 2.5 Flash
• 🧠 Design principle: LLMs handle language, deterministic Python owns anything exact
• ☁️ Google Cloud Certified Professional Cloud Architect
• 🎓 B.E. in ECE — The Oxford College of Engineering, Bangalore (2025)
• 📍 Bangalore, India · Open to Remote
• 📫 rajusivakumar453@gmail.com

Both reviewed and merged by the AVP of Neuro AI, Cognizant AI Lab.

AI / Multi-Agent Systems

Backend & APIs

Frontend

Databases & Cloud

6-agent Neuro SAN network extracting business rules from legacy COBOL, Java and PL/SQL for migration planning. Two CodedTools implement CodedTool with async_invoke() and handle deterministic work — parsing paragraphs, tracing CALL/IMPORT edges — because exact pattern matches cannot be left to a model. LLM agents handle interpretation, migration-risk assessment and spec generation. 84 unit tests cover the deterministic tools.

Python Neuro SAN HOCON CodedTool pytest COBOL Java PL/SQL

4-agent Neuro SAN network converting raw IT support email into schema-valid ServiceNow incident payloads. A deterministic PII-scrubbing CodedTool runs before any LLM call, so personal data never enters a model context. Pydantic v2 rejects out-of-range classifications at the schema boundary rather than emitting a malformed ticket.

Python Neuro SAN HOCON Pydantic v2 ServiceNow ITSM

Neuro-symbolic 3-way invoice matcher (PO ↔ Goods Receipt ↔ Invoice) for Source-to-Pay pipelines. Gemini 2.5 Flash reads unstructured invoices into typed Pydantic schemas; a pure Python rule engine executes every tolerance calculation. The LLM never performs financial arithmetic. Immutable SQLite audit trail with automated reconciliation reporting.

Python Gemini 2.5 Flash Pydantic SQLite Agentic Workflows

Production-grade fullstack AI platform — Next.js/TypeScript frontend + FastAPI backend — with Gemini 2.5 Flash, end-to-end RAG pipeline (ChromaDB), multi-agent orchestration, memory system, and Playwright browser automation for autonomous web task execution from natural language prompts.

Python FastAPI Next.js TypeScript Gemini 2.5 Flash ChromaDB RAG Playwright

"Building intelligent systems that solve real-world problems through AI innovation and automation."

⭐ Star my repos if you find the work interesting. Always open to collaborations, feedback, and AI engineering opportunities.