Hello, I’m Elad.

Coaches need useful performance data from underwater athletic footage. At Rubicon Robotics, I worked on the machine learning sub-models that help turn that video into measurable insights.

My internship combined software engineering with product strategy: connecting the analytics engine’s technical work to the information coaches need and the experience of using it.

I work on the lexicon team at the Humor Project, exploring the intersection of language and AI.

My broader role as Digital Product Manager & AI Engineer spans AI memory architecture, retrieval pipelines, and evaluation tooling. I also built an iOS application using on-device machine learning for contextual text generation, working on model optimization and API architecture.

I built no-gears-typing for the moment you realize you typed a whole sentence in the wrong language. It detects the mismatch and corrects the text as you type.

The program uses Rust and Python to switch to the intended language on the fly. After a pause, I’m back to actively developing it.

A model can claim tool-calling support and still fail with the tools you actually use. I built Callprobe to test your own tool definitions against OpenAI-compatible endpoints.

The harness scores tool selection, schema validity, and argument accuracy separately. It also tests when a model should avoid calling a tool, and how performance changes as more tools enter the conversation. The goal is to make the failures understandable, beyond a single benchmark score.