DAVID ZHANG

ML + CS + Applied Math | Columbia Engineering

About

I enjoy building in spaces where a large part of the problem is figuring out what to solve, and finding the abstraction that collapses complicated problems into obvious ones. I want to own my code end-to-end and build products that change how people work.

I co-founded Cyris, where we're building a metacognitive memory layer for AI agents in pharmaceutical R&D, cutting token costs by over 60% versus frontier models. Before that, at NephroNovus, I built an automated pipeline for protein-based drug development, cutting manual testing time by over 90%. On the side, I've been rebuilding language models from scratch (MLPs through Transformers) to better understand how they work.

Outside of tech, I'm a former varsity tennis captain, pianist who's performed in front of thousands, and an avid Pokémon collector and investor (around 10x returns in the past 18 months). I do my best thinking outdoors with classical music playing in the background.

Education

  1. Columbia University

    May 2028

    B.S. in Computer Science, Minor in Applied Mathematics

    GPA: 4.07

    Relevant Coursework

    Machine LearningNatural Language ProcessingData Structures & AlgorithmsCompetitive ProgrammingLinear AlgebraDiscrete MathematicsProbability for Engineers

  2. West High School

    August 2021 - May 2025

    High School Diploma, International Baccalaureate (IB) Diploma

    GPA: 4.61 (Weighted), 3.99 (Unweighted) | SAT: 1560 | ACT: 35

Experience

  1. Stanford AI Laboratory

    May 2026 - Present

    Research Assistant

    • Contributing to torch_measure, a PyTorch-native toolkit for predictive evaluation of AI systems.
    • Building latent variable models that infer system capabilities from sparse benchmark observations to predict real-world deployment.

  2. Cyris

    August 2025 - Present

    Co-founder, CTO

    • Building the metacognitive reasoning memory for AI agents by extracting failure-mode rules from historical case analyses, storing them in a sparse-retrieval rule store that is injected at inference time. Cyris predicts bypass pathway failures more accurately than frontier models while using 60% fewer tokens on the same reference context.
    • Currently being deployed for pharmaceutical R&D that predicts bypass pathway failures before phase 2 clinical trials.
    • Raised a pre-seed round for $100K at a post-money valuation of $2M, led by Convoi Ventures. Accepted for interview in Y Combinator’s Winter 2025 batch.

  3. NephroNovus

    June 2025 - August 2025

    Software Engineer Intern

    • Used Python and Jupyter Notebook to create an automated pipeline for the hit-to-lead and lead-optimization stages of protein-based drug development.
    • Shifted discovery from in-vitro to in-silico, utilizing NumPy and Pandas to process high throughput screening data, reducing manual testing time by over 90% in early-stage development.

  4. Mathnasium

    November 2024 - August 2025

    Mathematics Tutor

    • Simultaneously managed the progress of many students in a fast-paced environment as a tutor at Mathnasium.

Projects

MakeMore

  • Autoregressive, character-regressive language model that generates text similar to, but unique from, the input
  • Implemented ground-breaking research papers to build multilayered perceptrons up to Transformers from scractch
  • Added a wide range of techniques, such as manual backpropagation and batch normalization for the models
PythonPyTorchTensorFlow
GitHub →

MicroGradient

  • MicroGradient is an algorithm that performs gradient descent on data for binary classification
  • Recreates PyTorch's forward and backward propagation algorithms (currently with either ReLU or Tanh)
PythonNumPyPyTorch
GitHub →

Awards / Honors

  1. Harvard-MIT Mathematics Tournament, 10th Place General Individual

    2023
    • One of the largest and most prestigious high school math competitions in the world, drawing ~1,000 top students globally

  2. 5x American Invitational Mathematics Examination (AIME) Qualifier

    2020 - 2025
    • Scored 10 (Top 0.25% out of 300,000+ students internationally) in 2025
    • Scored 8 in 2023 and 2024
    • Missed USAJMO by less than one question in 2023

  3. American Mathematics Competition (AMC) Distinction Award (Top 5% Scorer)

    2020 - 2025
    • AMC 10 Distinction Award, 2021, 2022
    • AMC 12 Distinction Award, 2023, 2024

  4. USA Computing Olympiad (USACO), Silver Rank

    2022
    • Ranked Silver in USACO

  5. Utah State Mathematics Competition (USMC) Winner

    2022, 2023
    • Winner, 2022. Third Prize, 2023

  6. Sterling Scholar

    2024
    • Sterling Scholar in computer science in 2024

Leadership

  1. Mission Math Utah

    2022 - 2025

    Vice President

    • Managed and organized the largest high-school student run, non-profit math organization in Utah.
    • Reached out to 500+ local middle and elementary school students for awareness of competitive math. Raised and donated $15,000+ to local STEM organizations.

  2. Extended Learning Program (ELP) Math Club

    2021 - 2025

    President, Student Coach

    • Organized weekly meetings, coaching 30+ members to work through problems together.
    • Led the West High ELP Math team to win multiple chapter and state level awards at MathCounts and other contests.

  3. AP Calculus BC/AP Precalculus Teaching Assistant

    2023

    Teaching Assistant

    • Worked as a TA and grader for AP Calculus BC and AP Precalculus in 2023.
    • Helped students understand difficult concepts, graded homework and tests throughout the semester.

  4. West High Varsity Tennis Team

    2022 - 2025

    Captain, Varsity Player

    • Captain of the team in 2024, 2025. Varsity player since 2022.
    • Led 30+ members for fundraising and outreach events. Organized practices and coordinated tournaments for team members throughout the season

Skills

Languages

PythonJavaC++JavaScript/TypeScriptSQLHTML

Technologies

PyTorchTensorFlowNumPypandasReactNext.jsNode.jsFastAPIGitVercelAWS

Domains and Specializations

Machine LearningNatural Language ProcessingAgentic AIWorkflow AutomationFull-Stack Development