Shilan Jin

1. Data query and exploration using SQL and exploratory data analysis techniques.
2. Translate business/engineering problem into mathematical problems.
3. Select ML/AI algorithms to solve the mathematical problems effectively.
4. Dig insights from models to inform real-world complexity.

Evaluate complexity of projects, break down complex tasks and prioritize tasks.
Develop metrics to measure product/service quality from unstructured data by stochastic modeling, to enable effective decision makings.
Specialize in ML in continuous space, such as time series forecasting, functional PCA, kernel-based methods and non-parametric regression, and published three first-authored journals in top engineering journals.
Mentor a team of three junior researchers on ML methodologies of parametric and nonparametric regression and classification algorithms and variational Bayesian inference, ensuring their proficiency in applying these methodologies.
Predict image from image sequence using Generative AI (GANs and continuous normalizing flow) and transfer learning using LSTM.
Improved product quality by 50% and boosted production efficiency by 30%.
Was honored prominent research, highlighted in TAMU news [Link] and awarded NSF Fellowship (2021) with $60,000.
Segmented product qualities using clustering methods (DBSCAN and Gaussian Mixture Model) and classification methods (Decision Tree/Random Forest, and Support Vector Machine).
Developed a neural network framework for classification, i.e., Boosted Convolutional LSTM model, outperforming the conventional deep learning model in the imbalanced data regime.
Investigated into research of synthetic oversampling schemes, such as SMOTE, kernel Fisher discriminant analysis, and GANs (Generative AI), to enhance classification accuracy for minority class.
Implement and operationalize ML/AI models for practical use at research lab by deploying them as web services using Docker, Flask, and AWS.
Developed ML/AI algorithms outperform processing optimization benchmark methods by double-digit improvement in accuracy and efficiency.

Provided data-driven decision making schemes.
Query data from databases such as snowflake using SQL and process data using Hadoop.
Carried out exploratory data analysis with terabyte data using Tableau.
Formulated complex business problem into large-scale mixed integer nonlinear programming of thousands of decision variables and constraints.
Decomposed the complex mathematical programming into sub-problems to solve sequentially for efficient and accurate solutions.
Implemented mathematical programming models using SAS, Python, AWS, PySpark, and Gurobi.
Delivered presentations to senior executives and stakeholders.
Trained colleagues on the optimization platform.
Built recommender system using collaborative filtering and matrix factorization for data sponsorship.
Designed preliminary A/B testing to compare the recommender system with baseline system.
Formulated problem into large-scale dynamic programming, reinforcement learning.
Devised software testing methods (such as functional testing and simulation systems) to test system.