Yuda Li

Project engineer at Institut de Recerca en Energia de Catalunya (IREC) (2024-03 – Present)

• Analyzed the Spain electricity market and developed optimization models to integrate fuel cell generation, enabling participation in day-ahead and intraday market operations. Modeled market constraints, pricing mechanisms, and operational limits to ensure realistic and feasible dispatch strategies.
• Collected and analyzed public benchmark datasets on HVAC usage in residential and commercial buildings. Developed forecasting models to predict energy consumption and indoor temperature dynamics. Results of this work were published in the JCR-indexed journal Journal of Building Engineering.
• Implemented Mixed-Integer Linear Programming (MILP) models to optimize electric vehicle charging schedules, enabling demand-side flexibility and improved energy utilization.

Kaggle competition expert at Kaggle (2023-10 – Present)

• I am currently a Kaggle competition expert, holding one silver and one bronze medal. I am ranked 2,180th out of 204,792 Kaggle competitors worldwide (top 1%). I participated in the Kaggle "Detect Sleep States" featured competition hosted by the Child Mind Institute, where I implemented a Transformer + LSTM-based time series classification model. This model was trained on a dataset comprising over 100 million wrist-worn accelerometer samples, aimed at determining the probability of a person's sleep state, especially in detecting the onset of sleep periods and wake-up times. The model achieved an average precision score of 0.799, denoting its ability to predict sleep states within a 5-minute margin of error compared to ground truth events. This performance ranked 45th out of 1877 participating teams, earning a silver medal in the competition.
• I developed an ensemble model by combining a Temporal Fusion Transformer, gradient boosting models, and other neural network-based models to forecast energy generation and consumption of prosumers in the Baltic region. The model achieved a MAE score of 63.79, ranking 236th out of 2,731 participating teams and earning a bronze medal in the Enefit competition.

Project engineer at Universitat Politècnica de València & Universitat Oberta de Catalunya (2021-10 – 2023-10)

• Collaborated on various R&D projects, encompassing scientific articles and industry transfer initiatives, by providing expertise in data analysis, problem formulation, and algorithm design and implementation.
• Developed and deployed several efficient and interpretable ML regression models for an industrial partner, aimed at predicting economic losses from catastrophic events like hurricanes and earthquakes across various regions and countries. These models not only offer superior interpretability for integration into insurance policies but also deliver performance comparable to state-of-the-art ML algorithms.
• Developed auto fraud detection models for an insurance company using diverse ensembles of ML algorithms to identify fraudulent client claims, significantly outperforming the company's current baseline model.
• Conducted scientific research on integrating Discrete-event simulation from FlexSim with external Python optimization scripts to solve real-life complex optimaztion problems with uncertainty. This successful integration enabled the training of RL algorithms for solving storage allocation problems in warehouse settings, using FlexSim simulations as the training environment.
• Developed efficient metaheuristic algorithms in research projects to tackle a wide range of optimization problems in the literature, including stochastic vehicle routing problems, dynamic team orienteering problems, stochastic portfolio optimization problems, and stochastic warehouse storage allocation problems. The outcomes of these research efforts have been disseminated through publication in multiple academic papers and technical reports.

Investigation specialist at Amazon (2020-01 – 2020-10)

• Conducted holistic reviews and regulatory compliance checks to new and existing third party sellers on the Amazon platform in order to identify and prevent fraud sellers.
• Constantly exceeded weekly predetermined productivity targets and quality standards by more than 130% averagely.
• Identified several new fraud patterns which helped reduce task handling time.

PhD in Computer Science and Technology – University Carlos III of Madrid (2024-11)

MSc in Computational Engineering and Mathematics – University of Rovira i Virgili (2020-09 – 2021-09)

Bachelor in Aeronautical Management – Autonomous University of Barcelona (2015-09 – 2020-02)

Exchange program – Renmin University of China (2018-09 – 2019-06)