Austin, Texas · Founder and Principal AI Engineer
Ravinder Jilkapally
I founded AISOFT to help teams move AI from a working demonstration into software they can operate. I work with founders, CTOs, product leaders, and engineering teams across architecture, implementation, evaluation, and handoff.
What I work on
Twenty years of production engineering, applied to the current AI frontier. The throughline: systems that keep working after the demo — with evaluation built in, so the team that owns them can prove they work.
The operating range runs from agentic engineering — shipping with coding agents as part of the team — through small and local LLMs fine-tuned on proprietary data, to edge AI on hardware the customer owns. The company view is on the About page.
How I collaborate
- Architecture review — before a line is written, or when a system stalls.
- Build scope & production implementation — hands-on, code as the deliverable.
- Evals — ground truth, scoring, and the harness that proves which approach wins.
- Technical leadership & team enablement — working alongside your engineers, not above them.
- Owner notes — every engagement ends with documentation the next person can run with.
Evidence, not adjectives.
Shipped work, written up.
Undervolt
Urban infrastructure intelligence over 2.3 million Austin construction permits, running on local inference. First place, NVIDIA DGX Spark Frontier Hackathon. Read the build story and the Jetson port.
RefereAI
Sports officiating AI — vision models watching real match footage. The honest account of what a VLM can and cannot judge is in this build story.
Studio Copilot
A local AI workspace for photographers — built because NDA work rules out cloud upload. Won Raw to Curated at GTC. The build story.
StudyPal & TXLookup
Consumer products, shipped and running — education tooling and Texas public-records lookup. Both listed with the rest of the portfolio on the Work page.
Working on something in this range?
Email, or book time directly. Small and hands-on by design.