I'm Sravan Kumar Yadav, an AI/ML engineer. I build systems that work in production for real users, with clear thinking, solid engineering, and results that matter. Outside day-to-day delivery, I write about what I learn while building and evaluating these systems.
17K+
Enterprise Documents Processed
90%
Faster RAG Response Time
80%
Less Manual Deck Effort
The Story
A little of the person behind the work.
I got into AI because it felt alive. The field was moving, and I did not want to watch from the outside. I wanted to build inside it.
That same energy shows up in how I work. I get curious, dig in, and I stay with a problem until it makes sense. What is broken, why it broke, what we can do next. Simple questions. They keep me honest.
When something finally clicks, I like to write it down. Outside work, I run, stay serious about fitness, and I travel when I can. More than anything, I want people to feel steady around my name: that I can be trusted, and that I know what I am doing.
The Journey
The work I am in now, and the path that led here.
Jul 2024 – Present
Accellor Software Pvt. Ltd. · Hyderabad
Build and ship production AI for enterprise clients — search, document work, and systems people can actually trust.
17K+
Enterprise documents searchable with grounded answers, so teams stop digging through files by hand.
90%
Faster RAG responses from better retrieval and caching — less wait for users, lower cost for the business.
The Toolkit
What I reach for in production AI systems.
Core practice
How I design and ship agent systems.
The Work
Systems I am proud to have built.
Agentic RAG over 17K+ enterprise documents with multi-agent orchestration on Azure.
17K+ documents · 90% faster responses
Multi-agent RFP-to-proposal drafts with HITL review and Service Bus jobs.
Faster proposal turnaround
Automated ingestion from SharePoint, ADI, and Blob Storage with Container App Jobs for embedding generation and delta/retry processing.
Delta + retry resilient jobs
Hybrid RAG unifying SharePoint, Salesforce, and internal knowledge bases so every compliance answer is grounded in verified sources.
Audit-ready grounding
Text/PDF to branded PowerPoint decks — 80% less manual effort.
80% less manual deck effort
Credentials
Verified credentials that back the work.
The Thinking
Notes on building applied AI.
Get In Touch
Building something interesting? Let us talk.
Open to AI/ML roles and freelance projects that are a good fit. I usually reply within a day.
