Hayden Leung is a Computer Science student at Columbia University (class of 2028; minors in Biomedical Engineering and Linguistics). He has published graph-analysis research (IEEE IRI 2023), co-authored the MINA computational-genomics paper (accepted to iscbAI 2026 for an oral presentation, with the MLCB submission under review), engineered real-time haptic feedback for medical VR simulations at FundamentalVR, and researched LLMs on genomic data for early cancer detection at the Centre for Novostics. He also co-founded ResearcherX and open-sources developer tooling and agentic-AI projects.
Experience
Teaching Assistant, ENGI 1006 (Intro to Computing for Engineers and Applied Scientists) · Columbia University
– Present · New York, NY
TA for Columbia's largest introductory computing course (~220 students), teaching scientific computing in Python.
Mentor ~220 students in Python scientific computing (NumPy, Pandas, Matplotlib, Scikit-learn) via weekly office hours and debugging support.
Grade assignments and data-driven projects to rigorous standards for code efficiency, readability, and good data-manipulation practice.
Junior Research Assistant · Centre for Novostics
– · Hong Kong
Applied large language models to genomic data for a novel early-cancer-detection method, and evaluated DNA sequencing technologies for liquid biopsy in a wet-lab setting.
Investigated the application of large language models to genomic data to develop a novel cancer-detection method.
Evaluated DNA sequencing technologies in a wet-lab setting for use in liquid biopsy for early cancer detection.
Summer Intern · FundamentalVR
· London, UK
Engineered real-time robotic-arm haptic feedback for needle-puncture surgical simulations in C#/Unity.
Built scalable VR simulations in C# and Unity with physically accurate force thresholds and positional constraints.
Applied object-oriented design to model precise interactions with dynamic 3D objects in a needle puncture/withdrawal simulation.
Summer Intern, Communications Technology · ASTRI (Hong Kong Applied Science and Technology Research Institute)
– · Hong Kong
Automated end-to-end testing of commercial 5G network software with shell scripts, eliminating manual test execution; elected an intern division leader and press-conference representative.
Selected from 3,000+ applicants as one of 90 interns, and the sole high-school student admitted.
Built a shell harness that stands up the 5G core and gNB across ARM and x86 and runs the full end-to-end test flow.
Elected 1 of 4 Division Leaders and chosen as 1 of 8 Intern Representatives at the annual ASTRI Press Conference.
Education
B.S. Computer Science (major); minors in Biomedical Engineering and Linguistics · Columbia University, School of Engineering and Applied Science (SEAS)
– (in progress (class of 2028)) · GPA 3.858/4.00
Coursework includes Data Structures, Advanced Programming, Discrete Math, AI, Computer Systems, Databases, Natural Language Processing, and Computational Genomics.
Dean's List (every semester)
Harrow Diploma, Physical Sciences (Sixth Form / A-Levels) · Harrow School (UK)
–
Academic Scholar
A-Levels: A* in Chemistry, Physics, Mathematics, and Further Mathematics (4x A*)
Projects
ResearcherX
AI-native document editor that maintains a live knowledge graph of every claim, entity, and source in a document and lints new writing against it, flagging contradictions as you type, with GraphRAG chat and multi-format ingestion.
Co-founder and builder (planning, ideas, coordination, and part of the build)
A CLI agent that generates a whole repository's documentation set (README, guides, API reference, and more) and verifies every claim against the real source with a deterministic-first pipeline. Published to PyPI as docagent-ai.
An autonomous software-engineering system: a deterministic supervisor loop dispatches disciplined Claude Code sessions, isolates each in a git worktree, gates on a verified done/blocked/needs-human signal, and opens PRs, working a backlog to completion.
Sole author
Python · Go · TypeScript · Claude Code · git worktrees
A Claude Code framework and CLI for defensible academic writing: every citation verified and full-text-accessible, every paraphrase scope-matched to its source, and the author's voice preserved instead of flattened into AI-flavored prose.
A staged Python pipeline that finds every S&P 500 peak-to-trough drawdown of at least 10% and, using point-in-time index membership rather than today's list, reports the constituents that held up best in each one.
A recommender for the cold-start literature-review problem: given one seed paper it scrapes a citation ego-network, models it as a Markov chain, and ranks neighbours by Personalized PageRank, with a written analysis of convergence and damping sensitivity.
MINA: linear probes reveal coding-sequence family signal in frozen DNA encoders up to protein composition
iscbAI 2026 (ISCB AI symposium), Bali · 2026 · Second author; owned the statistical-rigor, baselines, and homology-aware-split robustness workstream · Accepted to iscbAI 2026 for an oral presentation (Bali, Aug 2026); MLCB submission under review
2023 IEEE International Conference on Information Reuse and Integration for Data Science (IRI), Seattle · 2023 · Co-led a team of five; led the graph analysis and the first k-core visualization over all of Wikipedia; presented at the conference · Published
Awards
2nd place, Lion's Cage (Columbia Venture Capital & Entrepreneurship Group) — 2026. For ResearcherX; only undergraduate finalists
Finalist (Top 6), Millard Chan '99 Technology Challenge — 2026. For ResearcherX; only undergraduate finalists
Dean's List (every semester) — Columbia University SEAS
CREST Gold Award (Engineering Award Scheme) — 2023. 130-page paper on Wikipedia's hyperlink structure; led to the IEEE IRI 2023 publication
UK Chemistry Olympiad, Gold Award — Royal Society of Chemistry, 2024
FIRST Tech Challenge, Winning Alliance and Control Award (London Regionals) — 2023. Lead engineer and driver; ranked 7th nationally