San Francisco, California, United States
I am a maker with a desire to change the world through technology. I’m driven by tackling challenging problems and learning at the edge of my comfort zone. This could be programming an internet of things gesture interface with a Microsoft Kinect, building software to make self-driving cars see or designing and building the largest cardboard boat ever to set sail on the River Cam. Today I'm on a journey to build the future of software with natural language. Outside of work I keep mentally and physically fit by training for triathlons.
Wordware helps domain experts rapidly create language model powered applications without having to code. It’s a Notion-like interface where complex chains of prompts can be interleaved with tools, loops and logic to produce powerful products and workflows. We pivoted away from HeyDaily but the core team remained the same.
We used to help creators earn a steady income but in the face of AGI we dropped everything and came back to our roots in AI. My co-founder and I have been working in ML and building AI-oriented companies since 2014. With the help of HeyDaily platform our creators earned $600,000 and we got close to profitability. We however decided to pursue the larger dream and pivoted.
Leading a team of highly skilled software and research engineers utilizing a mix of classical computer vision and neural networks to accelerate the development and assurance of self-driving cars. Working with product managers and designers to develop a compelling product and leading the technical design and implementation. Presenting the technology and product that my team has created to potential investors and customers (at executive and technical specialist level) which contributed to a successful acquisition of Five AI by Bosch in Q1 2022. Contributing to the implementation across the codebase including Python (heavy use of Numpy and JAX), React (with TypeScript) and Terraform with Kubernetes on AWS. Creating a large number of innovations led to being the first named inventor on 5 patent applications and a named inventor on a number of others.
Working in the Annotation Tools Team I helped push the capabilities of Five's advanced 3D annotation tool to new levels. This involved full-stack software development mostly in Python, React, TypeScript and SQL whilst heavily utilising Docker. I'm proud of the instrumental role I played in bringing additional sensor data into the front-end with high-performance using Three.js and local disk caching which I presented at the inaugural JavaScript Exchange in London. Additionally, I am the first named inventor on several patent applications.
I was initially contracted to convert a bulky radio-controlled pan-tilt-zoom camera system used for filming sports matches to a Wi-Fi tablet controller. The system required a diverse set of skills to design and implement: ▪ A Raspberry Pi core that ran a TCP server to take control commands ▪ An Android app that displays the live stream and provides a real-time control interface ▪ I2C LANC controller design, electronics and software for camcorder control and communication ▪ GStreamer for low latency live video with a low-cost video capture card Working independently, I delivered a well-documented, complete and cost-effective system on schedule. Since the initial contract, I have been doing ongoing support and feature development including: ▪ Surface mount PCB design ▪ Communicating with manufacturers to get quotes and assembled PCBs ▪ Live streaming video content to other devices/the internet over RTSP/RTMP
During this internship I created a tool to rapidly accelerate the annotation of lane segmentation data through SfM-based reconstruction and few-click human touch up via a web app. The platform that I built through my own initiative rapidly grew into a core team in the company generating 3D annotation data for all the deep learning in the company