Praveen — AI full-stack engineer
AI / ML · Full stack · Backend · Data · Frontend · DevOps
AI full-stack engineer, 5+ years. Building AI products end to end, from the first model to production.
NEWZera — my AI buddy for the Mac notch01 / about
Five years between a model and the moment someone uses it.
I design and build AI products end to end: real-time voice agents, LLM pipelines that know where to look, and the serverless infrastructure that keeps them fast. Research in, production out.
- Years building
- 5+
- Products shipped
- 07
- Industries
- 07
02 / stack · memory map
What I build with
StackINTELLIGENCE
LLM agents, retrieval and voice — models wired into products that answer back.
SERVICES
Fast APIs and the interfaces on top of them, from Go microservices to Next.js.
DATA
ETL pipelines, media processing and the stores that make embeddings useful.
CLOUD
Infrastructure as code, containers and serverless — shipped, observable, repeatable.
03 / selected work
Built to be used.
HOVER to lift · DRAG to pick up · FLICK to throw
DROP IN READER or CLICK to run · ESC eject
✦ / featured app · built and shipped by me
Zera
An AI buddy that lives in the Mac notch.
A native macOS app I design, build and ship on my own. Zera hangs from the notch and keeps a developer's day in one place: the Mac's own vitals, Claude Code sessions she can approve and reply to, fresh downloads ready to drop, tasks and timesheets written from your commits, and the clipboard, pull requests and calendar in between.
- Swift
- AppKit
- Core Animation
- IOKit
- Claude Code
- GitHub API
- Dependencies
- 0
- Tests
- 171
- macOS
- 13+
04 / notes · tail -f ~/notes.log
Field notes
Notes- [2026-09-24]#003Run MiniMax H3 locally on an RTX 5090 or RTX PRO 6000
- [2026-09-23]#002Run Qwen-Image 2.1 locally on an RTX 5090 or RTX PRO 6000
- [2026-09-22]#001Run FLUX.2 [dev] locally on an RTX 5090 or RTX PRO 6000
- [in the works]#···Serve an open LLM with vLLM on an RTX PRO 6000
- [in the works]#···Train a FLUX.2 [klein] LoRA on one RTX 5090
05 / connect