Shangar Arivazhagan (Keys)
Shangar Arivazhagan (Keys)
AI & Software Engineering Technical Lead
Python · AI Systems · LLMs · Backend Engineering

I'm Shangar (Keys), an AI and Software Engineering Technical Lead based in Coimbatore, Tamil Nadu.
I focus on building backend systems, agentic AI solutions that work reliably at scale.
This wiki is my digital portfolio where I document projects, technical decisions and learnings, and
experiments across AI and backend systems.
About Me
I lead technical delivery across AI and Backend projects, working closely with teams to design systems that solves customer pain points in the field of T&M and Enterprise Solutions. My day-to-day spans agentic workflows, API design, and keeping production Python services healthy and maintainable.
Outside of work, I spend time exploring AI advancements, working on personal projects, and writing up what I learn here.
Connect with Me
Tech Focus
I work across the full stack of modern AI and backend engineering. Here is where most of my depth sits right now.
AI Systems and Agent Engineering Building agentic workflows that go beyond simple chat: tool use, multi-step reasoning, structured planning, and MCP-based integrations. I care about making agents reliable, observable and evaluatable.
LLM Engineering RAG pipeline design, retrieval quality, structured outputs, prompt engineering and context engineering, and evaluation strategies. I work across OpenAI, Anthropic Claude, and Google Gemini depending on the use case.
Backend Engineering
Async Python with asyncio and httpx, FastAPI service design, event-driven patterns, Clean Code and clean domain
modeling with Pydantic and SQLAlchemy. I lean heavily on typed, tested, well-linted code as a baseline.
AI Infrastructure Vector database management with Qdrant, LLM observability, cost tracking, and evaluation with TruLens and custom harnesses. Getting the plumbing right matters as much as the models.
AI Tools and LLMs
LLM Providers: OpenAI, Anthropic (Claude), Google Gemini, Ollama, Hugging Face
Frameworks: LlamaIndex, LangChain & LangGraph (Intermediate)
Vector Databases: Qdrant
Evaluation: TruLens, custom eval harnesses
Embeddings: OpenAI, Ollama, Hugging Face
AI Coding: Claude Code, Antigravity, GitHub Copilot
Python Stack
Web Frameworks: FastAPI, Django
Supporting Libraries: Django REST Framework, Pydantic v2, SQLAlchemy, Alembic, httpx, asyncio
Dependency Management: uv, Poetry (Not Recommending for new projects)
Code Quality: ruff, mypy, pytest
Development and Tooling
Containerisation: Docker, Docker Compose
IDEs: VS Code (Professional) and Antigravity (Personal)
Version Control: GitHub, Bitbucket, Azure DevOps
Project Management: Azure DevOps, Jira
Documentation Platforms: Confluence, Azure DevOps Wiki