Abhuday Mishra
I build data pipelines and RAG systems that have to keep working for 100+ tenants, long after the demo is over.
- CRIS 2.0 ingestion pipeline: sources, validate, normalise, score, serve.
- WMP Copilot RAG pipeline: documents, chunk, embed, retrieve, answer.
Work
Abhuday Mishra (yes, he's referring to himself in the third person, and yes, he knows it's weird) is a Software Engineer at AiDASH, where he builds distributed systems, where the work happens to matter — helping utility companies predict outages, manage wildfire risk, and keep the lights on before the storm hits. His work sits at the intersection of backend engineering and AI infrastructure — most recently shipping a production RAG system that lets utility companies query wildfire mitigation documents across 100+ tenants, and a distributed ingestion pipeline that processes 320k+ records an hour powering climate risk predictions for enterprise clients.
When he's not wrangling Java, Spring Boot, and the occasional Python microservice, he's either chasing a shuttlecock on the badminton court, shooting hoops, or plucking away on his guitar. Music and sport — his two reliable escapes from a day of thinking about throughput and token costs.
He also writes about the unglamorous reality of building AI systems in production — evaluation frameworks, operational failure modes, and the decisions that actually move outcomes. Because the interesting part of RAG isn't the demo. It's everything that breaks between the demo and production.
Latest report
Building a RAG System: The Decisions That Actually Mattered
The settings that moved quality and cost in a production RAG system, from parsers and chunking to the orchestration layer, plus the gotchas hiding in a managed service’s defaults.
Read the report →Experience
Skills
Volunteer Experience
I ❤
On the web
Get in touch › abhuday.mishra@hotmail.com