At SHL (India), I built Chacha, an internal AI teammate for engineering and production support. It turns Jira tickets or alerts into an RCA report, a Teams alert, and a pull request, cutting root-cause analysis on a live incident from over two days to under five minutes.
I re-architected a core scoring flow in TCVPL, reducing typical runtime from two minutes to five seconds and worst-case runtime from three hours to 14 minutes. I also built GoMedia, a content-aware media package and service that cut conversion time to about 20 seconds per file and saved an estimated 194,000 compute hours annually.
My work at SHL has also included building SOS and the scoring pipeline for FreeSpeech, and replacing manual DBA query execution with a Liquibase-based pipeline and dashboard. I drove design for an upcoming voice AI assessment product and ran Cartesia and Langfuse proofs of concept to assess capacity needs.
