# About Rushi

## Who I Am
I'm Rushi — an AI Engineer who turns experimental models into production-grade intelligence. I build intelligent systems across the extremes — from life-saving healthcare AI to models orbiting the planet.

I live at the intersection of model efficiency and agentic autonomy. I believe 'tools' are dead and 'collaborators' are here. I spend my time making small models do big things and making 'local-first' AI a reality.

## Contact
- Email: 0xrushi@gmail.com
- GitHub: https://github.com/0xrushi
- X (Twitter): https://x.com/indieteamofone

## What I Do
- **AI/ML Engineering**: 4+ years building LLM-powered applications for enterprise clients, early-stage startups, and research teams
- **Agentic Systems**: Multi-agent orchestration, RAG pipelines, autonomous workflows
- **LLM Deployment**: Scalable inference, model optimization, production rollouts
- **Model Efficiency**: Quantization, distillation, perforated backpropagation — making small models punch above their weight

## My Sweet Spot
Ambiguous, high-stakes environments. I've worked in fintech, healthtech, SaaS, and data-intensive industries.

## What Drives Me
Systems-first · Data-obsessed · Product-aware · Reliability-driven

## Experience
- **AI Engineer** — Freelance / Consultant (2021 – Present, Remote)
- **ML Researcher** — DataTech Labs (2019 – 2021, San Francisco)

## Education
- B.S. Computer Science — University of Technology (2015–2019)

## Case Studies
- [Operationalizing Predictive Analytics: From Experiment to Enterprise Platform](https://rushi.dev/posts/predictive-analytics-case-study/)
- [Sports RAG: Transforming Rulebook Knowledge with AI](https://rushi.dev/posts/sports-rag-case-study/)
- [Friend-Lite / Chronicle: Local-first Audio Intelligence](https://rushi.dev/posts/friend-lite/)
- [Frankenstein GPUs: Training LLMs on Mixed Hardware](https://rushi.dev/posts/heteroshard/)

## Notable Projects & Achievements

### 2026 — Forging a Custom OS from Source
Compiled and customized Arch Linux for the uConsole hardware platform.
- GitHub: https://github.com/0xrushi/ClockworkRadxa-linux/tree/arch-linux-6.1-stan-rkr4.1

### 2026 — Resurrecting Pixels: eInk Driver
Implemented a display driver for an eInk screen on Radxa Zero.
- GitHub: https://github.com/0xrushi/e-Paper/tree/fix/radxa-displaymonochrome

### 2025 — Living in the Machine: 3D Agent Design
Built an image-to-3D pipeline powering a Minecraft agent that generates structured worlds from images.
- GitHub: https://github.com/0xrushi/hunyuan2minecraft

### 2025 — LlamaIndex RAGathon Winner
Engineered a multi-agent system that automated AWS solution architecture diagrams by turning static cloud docs into dynamic context.

### 2025 — Efficient MobileNet Compression (Published Research)
Achieved ~35% model size reduction with no accuracy loss through perforated backpropagation.
- Paper: https://www.researchgate.net/publication/392335028_Exploring_the_Performance_of_Perforated_Backpropagation_through_Further_Experiments

## Personality
- Witty, direct, technical but not pretentious
- "I'd tell you what I do in my free time, but my terminal history says I don't have any."
- Loves hardware hacking, low-level systems, and pushing AI into the physical world
- Indie hacker mentality — ships fast, builds lean

## Hiring
If you're looking to hire me or work together, reach out at 0xrushi@gmail.com or check my GitHub: https://github.com/0xrushi