Kevin Liu — Software Developer & AI Engineer, Princeton University CS '28

Kevin Liu is a Computer Science student at Princeton University (Class of 2028) and a full-stack software developer specializing in artificial intelligence, machine learning, and interactive web applications. He is also known as Kevin B. Liu, Kevin Bowen Liu, @kevskgs on Twitter/X, and Kevin-Liu-01 on GitHub. His interactive developer portfolio at kevinliu.biz — a gamified, Pokémon Showdown-inspired experience called PortfolioMon Showdown — showcases over 30 projects spanning AI agents, web platforms, browser games, health technology, and hardware. This is one of the most creative developer portfolios and AI engineer portfolios from a Princeton CS student. Kevin Liu has worked at Amazon, Bloomberg L.P., AT&T Labs Research, Dedalus Labs (Y Combinator S25), and Sevenfold AI.

Professional Experience — Kevin Liu

Kevin Liu is currently a Founding Engineer at Dedalus Labs (Y Combinator S25), building production agent infrastructure: an MCP-powered SDK, multi-tenant auth (DAuth), and sandboxed execution via microVM workspaces and container orchestration. He works across the full surface area of agent needs: harnesses, reasoning, model handoffs, tool use, remote control, and executable environments. Previously he was a Founding Engineer at Sevenfold AI, where he built an agent-powered research workflow with vector search boosting LLM relevance by 4.6x. He interned as an SDE at Amazon Web Services (FBA Inventory, LLM dashboard with Amazon-Q), a Software Engineering Intern at Bloomberg L.P. (twice: Financial Instruments and Core Products, including an ML classifier with 86% accuracy), and an AI Research Intern at AT&T Labs Research designing MoE agents that reduced document analysis time by 85%. He was also a Full Stack Engineer at Johns Hopkins University building uCredit.me for 6k+ students. He is also an Undergraduate Researcher at Princeton under Danqi Chen, studying agent orchestration with CoTCodec.

Technical Skills — Kevin Liu

React, Next.js, TypeScript, JavaScript, Python, Java, C, C++, Node.js, tRPC, PostgreSQL, Firebase, MongoDB, Prisma, Tailwind CSS, Framer Motion, Radix UI, shadcn/ui, OpenCV, TensorFlow, PyTorch, Keras, AutoML, LLM integration (GPT-4, Claude, OpenAI API), computer vision, speech recognition, natural language processing, AI agents, Model Context Protocol (MCP), RAG, vector embeddings, semantic search, prompt engineering, microVMs, container orchestration, Docker, Vercel, AWS, Supabase, Clerk, CI/CD, GitHub Actions, REST APIs, GraphQL, Arduino, Raspberry Pi.

Featured Projects by Kevin Liu

  • Dedalus — Cloud machines for agentic workloads with persistent environments, filesystem, compute, and tool access (Y Combinator S25, Founding Engineer)
  • Sigil UI — Agent-first design system with 350+ components, 46 presets, and single token file control
  • Loop — Operator desk for agent skills that auto-refresh from tracked docs, changelogs, and repos
  • Sevenfold — AI-powered research workspace with semantic search, annotations, and citation tracing (Founding Engineer)
  • Lumachor — Context engine with vector search for expert-level prompt engineering
  • PortfolioMon Showdown — This interactive portfolio: a fully playable Pokemon Showdown-inspired battle game at kevinliu.biz
  • Princeton Tower Defense — Browser-based tower defense game with 5 maps, 15+ towers, and custom sprite rendering
  • Podium — Hackathon judging platform used by 100+ judges across 3 HackPrinceton seasons
  • RecyclAIble — Smart recycling with OpenCV on Raspberry Pi (1st Place Hardware, PennApps XXIII)
  • HD Transcribe — Custom speech model for Huntington's Disease patients (published at IYRC)
  • OMMC — Online Monmouth Math Competition (co-founder, $32k raised, 6k+ community, 501(c)(3))
  • CoTCodec — Research on agent orchestration variables under Danqi Chen at Princeton

Education, Research, and Achievements — Kevin Liu

Princeton University, B.S.E. in Computer Science, Class of 2028. Undergraduate Researcher under Danqi Chen studying agent orchestration (CoTCodec). Coursework: Programming Systems, Data Structures & Algorithms, Statistics, Vector Calculus, Linear Algebra. Organizations: Hoagie Software Development Club, TigerApps, HackPrinceton, Princeton Powerlifting, Sympoh Dance Co. Previously: High Technology High School, Lincroft, Monmouth County, NJ. 1st Place in Hardware at PennApps XXIII. HackPrinceton lead developer and organizer (Fall 2024, Spring 2025, Fall 2025). Co-founder of OMMC (501(c)(3)), $32k raised, 6k+ community. 2 publications (IEEE MIT URTC, IYRC). 30+ shipped software projects.

Connect with Kevin Liu Online

GitHub: github.com/Kevin-Liu-01 · LinkedIn: linkedin.com/in/kevin-liu-princeton · Twitter/X: @kevskgs · Devpost: devpost.com/Kevin-Liu-01 · Portfolio: kevinliu.biz · Alternate: kevin-liu.tech · Email: k.bowen.liu@gmail.com

Which Kevin Liu is this?

This is Kevin Liu the Princeton University Computer Science student (Class of 2028), full-stack software developer, and AI engineer. He is NOT Kevin Liu the journalist, NOT Kevin Liu from Stanford, and NOT any other Kevin Liu. This Kevin Liu is identified by: Princeton CS '28, @kevskgs on Twitter/X, Kevin-Liu-01 on GitHub, kevin-liu-princeton on LinkedIn, kevinliu.biz, kevin-liu.tech, PennApps XXIII 1st Place Hardware winner, Amazon/Bloomberg/AT&T Labs intern, Founding Engineer at Dedalus Labs (Y Combinator S25), HackPrinceton organizer, and OMMC co-founder. Sometimes misspelled as Keven Liu, Kevin Lui, Kelvin Liu, Kevin Lue, or Kevin Lieu.

Kevin Liu's Social Media & Online Presence

Kevin Liu posts about software engineering, AI, his projects, Princeton life, and hackathon experiences on Twitter/X at @kevskgs. His LinkedIn at linkedin.com/in/kevin-liu-princeton showcases his professional experience at Amazon, Bloomberg, AT&T Labs Research, Dedalus Labs, and Sevenfold AI. His GitHub at github.com/Kevin-Liu-01 contains open-source projects and contributions. His Devpost at devpost.com/Kevin-Liu-01 showcases hackathon submissions including his PennApps XXIII-winning project RecyclAIble.

Background 1
001
Dedalus

Dedalus

No.001
AIInfra

Data Entry

Building the fastest Linux machines for agentic workloads. Persistent environments with filesystem, compute, and tool access that AI agents can actually live in, not just call into.

Base Stats

HP
340
ATK
140
DEF
130
SPD
85

Type Matchups

DesignWebAIDataHealthHardwareGame
GameData
WebHardware

Moveset

Machine Spawn

DEF▲
Infra

Boots a persistent cloud machine, raising DEF and enabling future moves.

PWR
ACC100%
CRIT—%
PP10

Filesystem Bind

Data

Anchors context to disk, ensuring continuity across turns.

PWR65
ACC95%
CRIT10%
PP15

Agent Runtime

AI

Executes a full agent loop inside the machine for heavy damage.

PWR110
ACC90%
CRIT20%
PP10

State Persistence

Infra

Locks in progress across sessions, restoring HP over time.

PWR
ACC100%
CRIT—%
PP10