At Kayak Ventures, we provide angel investment and startup advising services to innovative pre-seed and seed-stage companies.Managed by wife-and-husband duo Julia Lam and Andrew Imm, we aim to back great entrepreneurs building industry-defining companies in the tech and climate spaces.More at: https://kayak.ventures/Invest alongside us at: https://angel.co/s/julia-lam/d3Byp

From 2019 to 2021, a team of incredibly talented engineers rebuilt all of facebook.com from the ground up, leveraging the full power of our popular web technologies like React and Relay to create a faster, more app-like user experience. I supported multiple pillars of this team as an Engineering Manager, and continued to lead further integrations into the Facebook product after we shipped to the world.Over time, the underlying framework we built became the basis for the rest of Meta's web properties, including both established products like instagram.com, as well as new surfaces like threads.net and meta.ai. Realizing that we had created a flexible layer that abstracted away much of the web browser, we set out to do something most engineers would consider a pipe dream: sharing product code between web and native applications.For the last few years, we've been making this a reality. We've replicated that app framework in React Native, and have onboarded products that share 80-90% of their code between web and native. This has allowed those teams to reduce maintenance costs and accomplish more with fewer engineers, while shipping real native UI experiences on VR, Mobile, and Desktop.

Worked on Parse (www.parse.com) from Sept 2013 to Dec 2015, specializing in web performance and developer-friendly APIs.As the lead engineer on JavaScript client projects, I developed and launched open-source libraries like Parse+React (React bindings for applications built on Parse) and a rewritten version of the Parse JS SDK. I also researched ways to use observable interfaces in declarative programming environments like React.As the technical lead on the front end, I architected a new version of the Parse web site, leveraging technologies like React, Webpack, and Babel to build a system that dramatically improved performance, development time, and stability. With my focus on developer efficiency, it took our team of 4 individuals just 3 months to build out the entire site. We picked up some significant performance gains along the way, delivering all JS and CSS for the single-page app in a 300KB bundle, and reducing server loads by 40%. The project was later open sourced here: https://github.com/parseplatform/parse-dashboardFollowing that, I worked on designing novel ways to bring performance gains to Facebook's enormous web stack, adopting new technologies like Service Workers and offline execution.

As a transition to management, I built a team around the development of VR prototyping and design tools. After examining the workflow of AR and VR designers, we created a collection of tools, libraries, and plugins to give designers the ability to easily preview their designs in VR, collaborate and critique from a shared virtual space, and prototype multi-state interactions -- all built with standard web technologies.As part of this, we built a WebGL-powered UI library to match the rendering style of our first-party apps. We developed solutions to problems like using system fonts in WebGL, and built a custom React renderer that can draw React elements directly to WebGL framebuffers. Those UI layers were composited into full 3D scenes that could be interacted with and viewed on VR headsets using WebVR / WebXR. For collaboration and critique, we developed features like in-headset party calling via WebRTC, and Google Docs-style cursor sharing while multiple people browsed the same design document.

Worked with a multi-office group to develop React VR, an internal UI framework built on Facebook's React and React Native technologies that powers all of Oculus's first-party applications.React VR allows product teams to write React code, pull in elements from the React and JS ecosystems, and quickly iterate and build applications for VR headsets. It connects application code running from a JS engine with a custom GPU-accelerated rendering stack. It's designed to run on all Oculus platforms, and can target both mobile and desktop, rendering via OpenGL or DirectX. It's designed with VR inputs and constraints in mind, and powers the consumer experiences seen when you put on a Quest or Rift headset.

Bringing statistical analysis software into the 21st century.Stattery set out to solve a core problem with the entry-level statistics software used in many college classes: it's an oligopoly of packages that were designed by academics in the 80s and haven't been updated much since. They lack support on non-Windows systems, contain poor user interfaces, and require outdated licensing schemes. Stattery offered an alternative: as the "Google Docs of statistics software," Stattery let students get started with basic analysis on any web browser, saving their progress to the cloud, and providing straightforward workflows for the cases found in most introductory classes.Stattery was targeted at the use cases for individuals outside the field of statistics: students in introductory classes, and those in fields where some statistical analysis is necessary, but learning a language like R or SAS is overkill. It provided workflows for probability calculations, regression testing, hypothesis testing, various plots, and common distribution functions. Depending on the data entered, it was also capable of recognizing certain sample shapes and recommending an appropriate hypothesis test.Along the way, we developed some interesting technical hacks to get around the limitations of web applications at a time when Web Workers and weren't finalized or fully available. We designed a way to perform common matrix operations and SIMD calculations with WebGL shaders, placing the results in the color bytes of an offscreen canvas. Despite the cost of re-encoding to numeric values, we managed to get significant wins on supported browsers. The service also provided utilities to speed up tedious tasks, such as employing OCR technology to rapidly import data tables from paper sources like textbooks.

As an intern at Facebook, I worked closely with a new front end library that would later be released as React JS. I built out many of the earliest React UI components for facebook.com, and studied ideal developer patterns at a time when best practices for React were not yet understood. The internship culminated in the development of developer tooling that allowed engineers to live-code React layouts and learn about the library's single-direction data flow.