中文

Colar Wangedit

Graduate student at the University of Pennsylvania; builder of KitchenSurvivor and AgentConfig

Colar Wang is a graduate student and independent product builder based in the United States. He is enrolled in the Master of Science in Systems Engineering program at the University of Pennsylvania, and is the builder of the multimodal generative-AI consumer product KitchenSurvivor and the agent-configuration advisory tool AgentConfig.

Before his graduate studies Wang completed a Bachelor of Science in Financial Mathematics at the University of Nottingham and held three consecutive research roles in the Chinese financial sector, at CICC, CITIC Futures, and China Galaxy Securities. In the summer of 2026 he joined ByteDance / TikTok as an AI Product Operations intern.

Educationedit

Wang completed his secondary education in Shanghai before moving to the United Kingdom for his undergraduate studies. He received a Bachelor of Science with Honours in Financial Mathematics from the University of Nottingham in June 2025, graduating with an International Orientation Scholarship awarded to approximately the top five percent of international applicants.

In August 2025 he enrolled in the Master of Science in Systems Engineering program at the University of Pennsylvania's School of Engineering and Applied Science, where his coursework emphasizes applied machine learning, statistics for data science, simulation modeling, and marketing analytics. His expected graduation is August 2027.

Careeredit

For standalone product articles, see KitchenSurvivor and AgentConfig.

ByteDance / TikTok (2026)

In the summer of 2026 Wang joined ByteDance's TikTok division as an AI Product Operations intern on the Teen Safety team in San Jose, California. The role focuses on operationalizing LLM and classifier-based content moderation for the protection of minor users on the platform.

KitchenSurvivor (2025–present)

In November 2025 Wang founded KitchenSurvivor (Chinese: 云端小灶), a multimodal generative-AI consumer product aimed at the daily "grocery-to-dining" problem faced by international students. As Founder and Product Lead, he owns the full discovery-to-launch loop and elevated "recipe executability" — whether a generated recipe is physically cookable with declared ingredients — as the product's north-star metric. Detailed architecture and performance figures are documented in the main article.

China Galaxy Securities (2024)

In the summer of 2024 Wang served as a Quantitative Research Intern at China Galaxy Securities in Shanghai, where he productized a hybrid LSTM–XGBoost modeling framework for an automated trading system. A simulated backtest returned approximately 33 percent over the Q3 2024 window.

Earlier financial roles (2023–2024)

Before his quantitative research work, Wang held two internships in Chinese institutional finance: an Investment Bank Intern position at CICC in Shanghai (summer 2023), where he conducted semiconductor-market analysis and supported IPO prospectus preparation; and a Futures Department Intern position at CITIC Futures (winter 2023–2024), where he worked on Python-based strategy prototyping and investment-governance compliance.

Academic projectsedit

In addition to his shipped products, Wang has led two notable technical projects during his university studies.

Campus-Scale Img2GPS Localization (2025)

In a coursework project at the University of Pennsylvania between October and December 2025, Wang led a small team building a high-precision image-to-GPS geolocation system for campus-scale navigation. The project reported a 29.7 percent reduction in localization error and accelerated model convergence by a factor of 2.5, in part through a stage-wise adaptation strategy that reduced model parameter count by approximately 96 percent for deployment in resource-constrained environments. The full project report is available as a PDF.

AI-Driven Crypto Portfolio Optimizer (2024–2025)

In his final-year project at the University of Nottingham between October 2024 and March 2025, Wang led a small team building a deep-learning portfolio-allocation system covering 17 distinct crypto-assets. The system synthesized deep-learning return forecasts with volatility signals to drive automated rebalancing, and reported a roughly 90 percent reduction in forecasting error and a simulated Sharpe ratio of 1.3 under backtest. The full project report is available as a PDF.

See alsoedit