Richard (Yuxuan) Liao

Software Engineer

London, UK | +44 7919 706117 | a983974247@gmail.com

github.com/Shadowhusky | linkedin.com/in/yuxuan-liao

Software engineer building AI agents and business applications. Delivered agent-driven investment research workflows at Thema and independently released a terminal agent and cross-app automation tools. Previously worked with marketing and business development teams at ByteDance to turn campaign requirements into production software serving millions of daily active users.

Experience

ThemaLondon, UK

Front End DeveloperJan 2026 - Present
  • Built analyst workflows around AI research agents, connecting planning, streamed results, research artifacts and company shortlists for market research and investment diligence.
  • Made agent activity visible through tool-call inputs and results, with stop, retry and error handling. Added recovery so market creation resumes after page reloads or laptop sleep.
  • Delivered company screening through AI-generated table columns and rubric-based evaluations, with web research, streaming updates and recovery after refresh.
  • Developed a research and memo editor with version history, unsaved-edit recovery, embedded tables and nine chart types, enabling analysts to refine and export agent-generated research.
  • Added Playwright end-to-end coverage, offline MSW mocks and CI reporting to validate research and evaluation workflows before release.
  • Improved large-market navigation with virtualised lists and preloaded data; migrated charts to Recharts and removed approximately 30,000 lines of unused code.

ByteDanceShanghai, China

Front-End Developer IIAug 2021 - Feb 2024

Promoted from Front-End Developer in Oct 2022

  • Translated marketing and business development requirements into custom campaign pages and components, incorporating stakeholder feedback through rapid iterations.
  • Built reusable React components for an internal low-code platform, enabling marketing teams to assemble and customise coupon and campaign pages for high-traffic launches.
  • Led front-end delivery for the second of two consecutive Spring Festival campaigns, owning architecture, task allocation and complex features. Launched on schedule with no P0/P1 incidents during the event.
  • Supported Douyin marketing pages serving millions of daily active users through automated UI testing, CI/CD, release assurance and on-call troubleshooting across web and hybrid apps.

Alibaba GroupHangzhou, China

Front-End Engineer InternJun 2020 - Sep 2020
  • Developed government-facing applications in the DingTalk ecosystem, delivering three business iterations and five reusable components, including two cross-platform components built with Rax.
  • Built a VS Code extension integrating the internal low-code platform into developers' workflows and automated internationalisation tasks. Published technical guidance and received a return offer.
Richard Liao1

Selected projects

Tron | Open-source AI terminal agent

TypeScript, React, Electron, Node.js, xterm.js, node-pty

  • Built an LLM agent that plans multi-step tasks, executes commands in a real shell and uses output to choose subsequent actions and recover from errors, locally or over SSH.
  • Added command permissions, dangerous-command checks, loop guards and persistent sessions across cloud and local models. Shipped macOS, Windows and Linux installers with 300+ downloads across GitHub releases.

Recast | AI writing and automation across apps

TypeScript, Svelte, Tauri, Rust, Swift

  • Built a macOS writing utility with pipelines chaining AI rewrites, shell macros and agent steps; added explicit command approval, streaming output and automatic model fallback.
  • Released an iPhone companion on the App Store with keyboard, Share and Shortcuts integrations; shared prompt templates and provider behaviour across platforms.

Cipher Zero | AI deduction game

TypeScript, Next.js, React, Socket.IO, SQLite, Jev, Docker

  • Built and deployed a bilingual deduction game combining generated decks with Jev's structured judgments. Implemented solo and real-time multiplayer modes with server-controlled rules, persistent progress and reconnection.
  • Implemented evidence retrieval, confidence thresholds and model fallbacks for uncertain answers; preserved unresolved candidates when evidence was missing or contradictory.

Additional experience

University College LondonLondon, UK

Postgraduate Teaching Assistant, App EngineeringJan 2025 - Apr 2025
  • Guided students through Next.js application development, explaining architecture and debugging projects using Prisma, NextAuth and SQLite.

Education

University College London2024 - 2025

MSc Computer Graphics, Vision and Imaging | Distinction, 80%

Dissertation (86%): real-time gesture input on Meta Quest 3; snap detection AUC 0.967 with 156 ms latency.

University College London2018 - 2021

BSc Computer Science | First-Class Honours

Skills

Engineering: TypeScript, JavaScript, Python, React, Next.js, Node.js, SQL; Electron and Tauri

Agents and AI: LLM APIs, tool calling, prompt templates, streaming, model fallback, permissions, recovery, evidence retrieval

Delivery: Playwright, Vitest, MSW, GitHub Actions, Docker, Cloudflare Workers

Languages: English (fluent), Chinese (native)