# Akash Panchal

> AI systems engineer at Salesforce. Ten years shipping production software.

## Summary

Akash is an AI systems engineer with 10 years of experience across Salesforce, Amazon, and McAfee. He builds agent runtimes, voice agents, and independent products in TypeScript, Python, React, FastAPI, and AWS. Independent products stay on a low-touch cadence. They are live and still moving.

## What he builds

- Agentic products: multi-turn systems with skills, tools, memory, evaluation, and human approval points.
- Voice agents: low-latency, interruption-safe systems using LiveKit, WebRTC, VAD, streaming STT/TTS, and dependable tool execution.
- Apps: end-to-end products, from an unclear customer problem through UX and production implementation.

## Selected work

- [Content Studio](https://www.salesforce.com/marketing/) (Salesforce): Multi-turn agent workspace for Marketing Cloud. Harness primitives for threads, memory, skills, tools, and approvals.
- [OneCue](https://onecue.sh): Decision layer for coding agents. Surfaces the prior decision a change could break, with the reason, then stays silent when nothing applies.
- [AI SDK Patterns](https://ai-sdk-patterns.dev): Open-source TypeScript patterns for the AI SDK, a composer that turns them into a Next.js app, and a 16-chapter learn track.
- [VibeFrames](https://vibeframes.vercel.app): Agentic video workspace. A Mastra harness turns a brief into an editable Remotion composition with observable tool calls.
- [Evercall](https://evercall.app): AI call handler for Australian emergency plumbers. Captures missed calls, triages jobs, books follow-up, SMSes the owner.

## Machine-readable

- Portfolio: https://akashpanchal.com
- Structured profile: https://akashpanchal.com/profile.json
- Writing index: https://akashpanchal.com/llms.txt
- Résumé: https://akashpanchal.com/CV.pdf
- GitHub: https://github.com/akashp1712
- LinkedIn: https://www.linkedin.com/in/akashp1712