1 min
◆ About
I'm Rajan, and I build retrieval systems
Senior AI engineer with 3+ years building and operating production AI systems, currently leading a team of five and owning the stack end to end: document intelligence and RAG pipelines, multi-agent orchestration, vLLM serving on Kubernetes, and GPU cost engineering. Depth in high-throughput inference (vLLM, TGI, Triton) and parameter-efficient fine-tuning (QLoRA), delivered across AWS, Azure, GCP and DigitalOcean.
Most of my work sits at the seam between a language model and the documents it is supposed to reason over. That seam is where projects actually succeed or fail: how a page is segmented, what survives transcription, which retrieval strategy is used, and whether anyone notices when it quietly degrades.
Lately that has meant a two-stage document pipeline that separates layout detection from transcription, long-horizon extraction agents that stay coherent across hundreds of documents, and a temporal memory layer on Postgres and Qdrant whose in-memory index is 32× smaller after binary quantisation — with recall held up by rescoring rather than hope.
A good deal of it is unglamorous cost engineering: moving a serving path onto cheaper GPUs without losing throughput, autoscaling inference so it is not paid for while idle. The recurring lesson is that the interesting engineering is almost never in the model.
Why this site exists
Writing is how I make sure I actually understood something. The essays here are the long-form version of that; the series are for subjects that need more than one sitting; and the interview notes are what I revise from, published in case they are useful to someone else preparing.
If you are a researcher, a hiring manager, or someone who has run into the same problems — I would genuinely like to hear from you.
§ Writing