Md Ayan Arshad - AI Engineer
I build GenAI and Agentic AI systems for production. RAG pipelines, multi-tenant architectures, agentic workflows — the full stack from design to deployment.
At Softeon I joined as one of the first people on the Data Science team. Talked directly with domain experts and AVPs, took requirements from the customer support team, designed the architecture, and shipped with a small team. No playbook, figured it out end to end.
At Softeon I shipped two systems:
-
Multi-tenant RAG chatbot across 500+ enterprise clients. Tenant isolation at the vector DB layer, hybrid retrieval with cross-encoder reranking. Faithfulness pushed from 0.67 to 0.91. Unsupported claim rate below 4%.
-
AI-powered ticket resolution system for the WMS support division. Took requirements directly from the support team, built semantic retrieval over historical tickets and documentation, rolled it out end to end.
Outside of work I built a kapa.ai (YC S23) inspired multi-tenant RAG platform. 12+ retrieval pipeline combinations tested before settling on the final stack. RAGAS scores: 0.91 faithfulness, 0.95 context recall, 0.89 context precision. MCP server integrated with Claude Desktop. Full observability with Prometheus and Grafana. Fully documented and containerised with Docker.
I design systems keeping all the tradeoffs in mind and write about the decisions tutorials never cover, on Dev.to and LinkedIn.
Looking for a fast-paced startup where I can own problems end to end, talk to users, and ship for them directly.
I do one free RAG and Agents architecture review per month. 48 hours. I’ll find where your system breaks before your users do.