Experience archive

University of California, Los Angeles (UCLA)

M.S. in Computer Science

University of Southern California (USC)

B.S. in Computer Science · Magna Cum Laude · GPA 3.76/4.00

University of California, Santa Cruz (UCSC)

Computer Engineering · transferred · GPA 3.97/4.00

American Heritage School, Boca Delray

Secondary education · grades 11-12

Dulwich International High School Suzhou

Secondary education · grades 10-11

Liangzhu Anjilu Experimental Middle School

Middle school · grades 7-9

Zhejiang University Affiliated Primary School

Primary school · grades 1-6

Health Intelligence Lab (HAIL) · UCLA

Research with Prof. Yuzhe Yang

Researching multimodal foundation models for diabetes using CGM, insulin, meals, and activity; built patient-disjoint MetaboNet benchmarks for 30/60/120-minute forecasting.

AI Engineer Intern · PMAOS / Next Big Seek

Recommendation algorithms, multimodal inference, and robotics research planning

• Designed bidirectional user–video interest propagation for short-video recommendation, separating learned preference, retrieval scores, and diversity annotations.
• Allocated exposure through session, bridge, and long-term interest budgets, with fatigue controls that reduce repetition while preserving exploration.
• Designed an 11-language audio-visual affect pipeline with A/V offset rejection, speech PAD, evidence fusion, and a Q15 decision layer that can abstain or report conflicting modalities.
• Validated the recommendation work with 629 tests, strict Mypy across 438 source files, and 34 contract checks; the multimodal pipeline passed 139 Python tests and C build gates for Cortex-M3 and ARM926EJ-S.
• Screened sub-10 mm untethered piezoelectric robot architectures and specified a gated PZT-steel cantilever experiment for a 9 mm candidate. Work ended at experiment and procurement planning.

Co-Founder & CTO · CrAIg Capital

Los Angeles, CA

• Led product and engineering for an AI-assisted housing debt workflow, including data pipelines and automation.
• Set technical direction, hiring priorities, and delivery standards.

Full-Stack Engineer · Terlingua Threads

Remote

• Built and optimized Shopify pages, responsive layouts, and maintainable content structures.
• Developed Google Maps features and improved accessibility and SEO.

AI Engineer Intern · Piezolab Medical

Lake Forest, CA

• Built an LLM-assisted pipeline for more than 5,000 medical malpractice documents.
• Developed agent workflows, experiment checklists, and acceptance tests for vascular-care research.

Business Development Intern · Hangzhou Haoan Jianke Technology Co., Ltd.

Hangzhou, China

• Rotated through major departments and helped gather requirements for an internal platform built with NetEase CodeWave.
• Helped recruit engineers and plan the program structure and interface.

Investment Analyst Intern · GUDU Medical

Remote · MedTech investment

• Supported deal screening, market analysis, due diligence, financial modeling, and valuation work.

Application Engineer Intern · Synaptics

San Jose, CA

• Validated firmware for an internal product.
• Used an ESP32 and Alibaba Cloud to route I²C sensor data into MQTT topics for internal dashboards.

MedTokenBudget: Lesion-Preserving Token Routing for Dermoscopic Image Classification

Under review

• Introduces lesion-aware token routing under a target token budget, trained with mask supervision, budget curriculum learning, diversity regularization, and attention distillation.
• Audits how much ground-truth lesion evidence remains after routing on ISIC 2019.

Mass-Conservation Token Transformers for Auditable MS/MS Spectrum Retrieval

Under review at AAAI 2027 · preprint forthcoming

• Represents fragment-pair mass-conservation evidence as sparse tokens before spectrum ranking.
• Uses closest-mass and overlap-hard negatives, then compares rankings before and after the evidence tokens are removed.

Efficient Attention via Pre-Scoring

First author · advised by Prof. David P. Woodruff (CMU)

• Proposed a pre-scoring mechanism that selects informative key intervals before attention.
• Implemented Triton kernels and compared the approach with HyperAttention and LevAttention.

Game Theory in Companies

Advised by Prof. Ming Gu (UC Berkeley)

• Led a team that modeled game-release decisions as two-player normal-form games and used Python with Gambit to enumerate Nash equilibria.
• Li, S., Li, Z., & Liu, Z. (2023). Advances in Economics, Management and Political Sciences, 16(1), 357-366.

USC Drug Discovery Research

SARS-CoV-2 antiviral discovery with USC Prof. Yan Liu's group

• Benchmarked RDKit, Chemprop, and hybrid models across seven TDC datasets with scaffold splits.
• Ran RDKit importance and SHAP analysis; Extra Trees reached AUROC 90.0 ± 6.4 on SARSCoV2_3CLPro_Diamond.

Fridge Manager

Inventory-aware meal and grocery planning

• Built a mobile Next.js + Supabase product that combines inventory, leftovers, recipes, preferences, and time constraints.

AI Doctor WeChat

TCM clinical and teaching assistant

• Built workflows across a WeChat Mini Program, web workspace, and FastAPI backend for doctors, assistants, students, and patient intake.

AI Business Card Hub

iOS business card organizer

• Added vision-first parsing, OCR fallback, multi-photo capture, structured people/company records, bilingual review, and search.

CNN Deep Fake Face Detection

Keras image classification

• Trained a three-block CNN on more than 10,000 real and GAN-generated face images.

TradeViewer

React + Java transaction visualization

• Led an eight-person team, set objectives, assigned work, resolved conflicts, and integrated the frontend and backend.

Silicon Valley Company Tour

Noah Business School program · company visits and conversations with founders and investors

TIE Hub Co-Founder Speed Date · Volunteer

Venue setup, check-in, guidance, timekeeping, and event coordination

USC Badminton Club

Member · regular participation through junior year

TIE Hub Startup Hackathon

Five-day startup workshop · concept development, working demo, venture design, and pitch

Magna Cum Laude

University of Southern California

Viterbi Dean's List

University of Southern California

VEX Robotics Asia-Pacific Championship · Champion

High School Division

VEX Robotics Asia-Pacific Championship · Champion

Middle School Division

Noah Business School (Noah Holdings)

Entrepreneurship and executive program covering venture building, financing, and company visits

Tutor · Introduction to Machine Learning

UCSC · Prof. Lise Getoor

• Held office hours and discussion sessions on logistic regression, SVMs, tree models, and related topics.
• Co-designed Jupyter Notebook assignments and contributed to the coding autograder.

Assistant Coach · VEX Robotics & Olympiad in Informatics

Shanghai Bosidi Education

• Helped train teams in robot design and algorithmic problem solving, and supported scrimmages and events.