Hi, I'm Prajwal

An AI engineer based in California. I build AI agents and the developer tools around them, right now emberd, a runtime that runs untrusted AI-agent code inside isolated microVMs. Before this I spent three years at Opslyft on cloud cost and reliability systems, then a Master's on adversarial ML and Android malware detection. I'm open to full-time AI engineer, backend, and full-stack roles.

Now

Open to AI engineer and backend/full-stack roles with end-to-end ownership. California, remote, or hybrid.

Building emberd, a Firecracker-based sandbox runtime that runs AI-agent code inside isolated microVMs. Also maintaining QuackCode and Gitwise.

Reading: AI Systems Performance Engineering right now, and papers on AI agents.

Projects

Open Frame preview

Open Frame

Make slides, carousels, thumbnails, and OG images with React and a coding agent.

ReactTypeScriptAI Agents
BaitBench preview

BaitBench

Measures how often an LLM, writing a job application for you, falls for a prompt injection hidden in the job post.

Prompt InjectionLLM EvalsSecurity
emberd preview

emberd

Run untrusted AI-agent code in isolated microVMs you host yourself, inside your own network.

GoMicroVMFirecracker
codexpass preview

codexpass

Use your Codex login anywhere, no separate API key to set up.

GoCLIDeveloper Tools
All projects

Featured work

Re-Evaluating Android Malware Detection: Tabular Features, Vision Models, and Ensembles

Hosahalli Dayananda, P.; Chen, Z.

Compares tabular features, visual byte-plot images, and ensemble strategies for Android malware detection. Evaluates where modality choice changes the detection surface and where ensembles actually pay for themselves.

Writing

All posts

Components

TOC Minimap

A table of contents that sits on the edge of the page as a row of ticks you can scrub through.

All components

Talks

Data Transfer Cost Optimization

Rootconf 2022

Granular visibility into cloud data transfer costs using AWS CUR and VPC flow logs.

View talk

Simplifying Kubernetes Cost Visibility

Opslyft 2022

Webinar on reducing the complexity of understanding and attributing Kubernetes infrastructure costs.

Get in touch

Open to AI engineer and backend/full-stack roles, California, remote, or hybrid. Email is the fastest way to reach me.

Command Menu

Quick navigation and actions