A BIT 'BOUT MY JOURNEY
I'm an AI Engineer who believes the best systems come from obsessive attention to detail wrapped in healthy curiosity.
My journey began at an MNC, where I fell in love with the chaos-and-thrill of real production systems. Since then I've shipped end-to-end: a multi-agent fraud resolution system at Visa, serverless document & email automation pipelines at Transport for London, and self-service platforms at Aviva - always owning it from first sketch to live.
Lately I've been deep in agentic AI, RAG, and MCP - and I'm building my own agentic harness to experiment with smarter, adaptive automation. 👀 Don't forget to take a look at my Personal Projects section!
- 🕵️
Shipped a multi-agent fraud resolution system at Visa that cut manual case handling by 40% across 10k+ daily cases
- 💸
Cut LLM operating costs by 35% through intelligent model routing & semantic caching
- 🔎
Pushed RAG retrieval accuracy from 72% → 91% with a graph microservice that retrieves logical relationships, not just text matches
- ⚡
Automated 5,000+ emails/day at 95%+ classification accuracy for TfL
What drives me? Insatiable curiosity - and the belief that engineering should be a little bit exciting.
LangGraph
MCP
OpenAI
Claude
Azure OpenAI
Agent Orchestration
Function Calling
HITL
Multi-Agent Workflows
RAG/GraphRAG
Hybrid Retrieval
LangSmith
MLflow
LLM-as-Judge
Regression Testing
Experiment Frameworks
Latency/Cost Benchmarking
Prometheus
Grafana
Python
FastAPI
REST APIs
Microservices
PostgreSQL
pgvector
Neo4j
AWS
Docker
Kubernetes
ArgoCD
Temporal
Work
Check out some of my recent work - cool products I've been part of, built using the latest tech and libraries to create something awesome!
Built an agentic AI workflow that classifies transactions as fraud/non-fraud and automatically triggers downstream business actions like dispute creation and card blocking using tool-calling agents and a custom orchestrator.
See what I did
Transport for London (TfL)
Developed data processing workflows to extract information from emails and documents, enabling downstream teams to use the data effectively.
See what I did
Built and supported the internal IT Self-Service Portal used by Aviva employees to log support requests, access shared tools and resources, and request services. Worked primarily on support, bug fixes, database-related tasks, and small feature improvements to enhance portal reliability and performance.
See what I did
Personal Projects
Check out some of my recent work - cool products I've been part of, built using the latest tech and libraries to create something awesome!
AgentHarness is my own open-source framework for building and running LLM agents - a Python runtime with durable orchestration (async event bus, scheduler with backlog + cancellation), an async job queue with multi-worker scaling, and dynamic on-demand skill loading on top of an MCP-backed tool registry.
See what I did
Multi-Tenant Agentic RAG is a Python/FastAPI implementation of a multi-tenant Retrieval-Augmented Generation system split into three services - ingestion (parsing, OCR, chunking), embedding (Voyage AI + Pinecone, one namespace per tenant), and retrieval (query validation, vector search, grounded answers with citations).
See what I did
MCP with Auth is an MCP server that exposes a web search tool (backed by Tavily) over the MCP Streamable HTTP transport, wrapped in a FastAPI server and secured with OAuth 2.0 bearer-token authentication via Scalekit - implementing the MCP OAuth 2.0 protected resource metadata flow so standards-compliant clients auto-discover the authorization server and scopes.
See what I did
Client's love
Had the opportunity to work with Mann during my time at Visa on Dispute Automation. Mann's critical thinking and ability to own the deliverables was impeccable.
Shashikant Dwivedi
Visa
"I had the pleasure of working closely with Mann on the TfL project. He played a key role in designing and building scalable serverless document and email-processing workflows on AWS that significantly improved our operational efficiency. Mann has a calm, analytical approach to problem-solving — whether it was architecting reliable OCR pipelines, smoothing complex automation flows, or troubleshooting production issues, he consistently delivered high-quality solutions."
Mahuya Mondal
Transport for London (TfL)
"We worked with Mann at Aviva on several backend and automation initiatives. He consistently delivered high-quality solutions, approached problems thoughtfully, and helped improve our team's engineering practices. Mann was always willing to collaborate and support others, bringing both skill and positivity to the team. We highly recommend him to any engineering team."
IT Automation & Engineering Team, Aviva
Aviva
Lets build beautiful,
meaningful things together!
+91 9724663588