Prototyping open-source web3 metaverse products at various hackathons and winning some prizes along the way.* At Eth New York in June, my team won first place out of more than 100 submissions to the Protocol Labs/Filecoin track for W3B_W@Y, a protocol that allows 3D artists to save virtual worlds to decentralized storage and tokenize them on the blockchain.* At HackFS in July, our team was selected as a top 10 finalist out of 197 teams for w@y_p01nt, a user interface and protocol that enables 3D artists to build and mint a decentralized network of virtual worlds as ownable ERC721 tokens.* At Eth San Francisco in July, our team won the Best Metaverse or Gaming award from both Polygon and SKALE for interX, a user interface and protocol that allows game developers to build on-chain interactive and interoperable metaverse asset NFTs that can exist beyond the border of a single game.

* Arium is a browser-based, real-time collaborative platform for NFT creators and collectors to build 3D virtual worlds and hold engaging and fun live events around their nfts to engage their communities. It features spatial video and audio chat, providing a connected and serendipitous social experience for users.* The platform spread virally through the cryptoart community and was used repeatedly by some of the most influential figures and communities in the space, including Hackatao, Vertical Crypto Art, Whale Community, MoCDA, and A.L. Crego.* As CTO and co-founder, I built the tech stack from the ground-up. The browser-based collaborative virtual world builder was made with react and three.js for the 3d scene composition and rendering. Firebase was used to store and synchronize content across user sessions.* For peer-to-peer webrtc video and audio communication, we built a horizontally scalable distributed server cloud and Google Cloud, node.js and MediaSoup. The platform was designed to be highly performant and scalable, with much of the browser based computation offloaded to webworkers, and we were able to support up to 100 concurrent users in a single world at 60 frames per second.

* Worked with the R&D team to investigate computer vision techniques for extrapolating 3d pose of athletes at live sporting events, using multiple camera, burst photography, and machine learning models. * Build a framework in python that used OpenCV and SciPy determine camera position from point marker on screen.* Evaluated different pose estimation models for accuracy, performance, and developer friendliness.* Built human-in-the loop system for ability to correct inaccuracies in pose estimation results.* Built a machine learning processing pipeline in Kubernetes and Google Cloud to quickly process captured photography quickly to turn it into 3d pose data for journalistic reporting.

Resident Researcher at the Interactive Telecommunication Program at NYU. Areas of work included using real-time computer vision techniques in interactive contexts (sports, art installations), smart sensing using low-powered ml powered edge computing hardware, and audio and video streaming in virtual spaces.

* Built PoseNet and BodyPix, browser-based machine learning models for real-time human pose estimation and body-part segmentation, respectively, using internal research from the Google Machine Perception team. Collaborated closely with the researchers, the Google Creative Lab and the Tensorflow.js team to bring these models to life.* Continued to work with the research team beyond the scope of the initial contract to improve the accuracy and capabilities of these models.* Published widely-shared blog posts about both projects, making the research and capabilities accessible and understandable to a wider audience.

* Lead full-stack engineer for the Pursuit by Equinox, a gamified spin class experience that uses over 500,000 points of real time bike data per class to visualize performance and drive competition.* Lead a team to of re-build the platform to make it more reliable and easier to develop against. We re-built it in react, redux, node.js, and three.js, with state centralized on the servers and pushed in a stream to the clients. Our efforts took the time it take to release a new game from 2 months to 3 weeks.* The class was deployed to 40 physical Equinox locations around the us, with 20,000 riders per month.Spoke at the NYC Node.js meetup about the work:https://www.youtube.com/watch?v=zo39p-30arg&feature=youtu.be

Full stack engineer building browser-based interactive applications.Worked at various companies. Details left out here to keep the LinkedIn profile concise.