- Work on Embedding representations for multi-target Cross-Domain Recommendation.- Concentrate on multi-target CDR models applied to multiple types of items with an overlap on the set of users, and address through embedding-based transfer approaches, from classic embeddings creation based on matrix factorizations to the most modern approaches using deep learning and transfer learning.- Overview of single-domain recommender systems. Special attention to CF and DL approaches.- Adapt to work in a multi-target Cross-Domain Recommendation context.- Comparison analysis of the different approaches and conduct empirical study on various techniques, like transfer learning (BERT4Rec, ShopperBERT, etc.), and multi-task learning.

• Assist in building an event-driven backtest system in Python to perform strategy backtesting, parameters fine-tuning, andprovide a realistic simulations environment that mirrors live deployment of a trading strategy.• Develop statistical, time series models, and machine learning models including varying parameters regression, random forest, xgboost, to provide better and robust models for asset prices forecasts.• Collect data from internal API, and external resources (Bloomberg, Thompson Reuters), and analyze data using Pandas library, and statistical tools in Python and R in alpha research project.• Implement cloud computing solutions in AWS, Azure, and deploy deep learning models for sentiment analysis on financial data.• Develop the risk models and the estimation and calibration procedures in company financial model development framework (e.g. Risk management using Hidden Markov Model, and VaR/CVaR estimation on portfolio)• Carry out unit testing on Preprocessing functions, and financial models to ensure all functions and models are bug free and work as expected.

• Built an automated bank reconciliation algorithm using Python that automates the end-to-end bank reconciliation process such as checking and matching transactions, and producing monthly bank reconciliation statement. The algorithm is currently still in use and has significantly reduced the time spent on daily reconciliation process. • Wrote algorithm that automates activity based costing calculation for over millions of existing contracts, and analyze contracts that have big P&L impacts to the company, and recommend cost reduction suggestions to senior management team. • Collaborate with airport control team in building a Long Short-term Memory (LSTM) model that forecasts airport passenger flow using Pytorch to help promote crowd control. This resulted in a global 40% decrease in customer complaints.

• Managed entire company depreciation accounts and ensure all capitalized items are in compliance with company standards, analyze all capitalized assets and forecast capital budget using ARIMA model to produce financial reports that accurately reflect the business’s value. • Built an automated cost model script for each capitalized assets (Sort/Assembly Equipments) in Excel to help decision making for the senior management level. This has led to significant budget savings of $5 million yearly. • Led in a team with mixed finance and engineering departments for detailed analytical and technical review of business operational process including the depreciation policy project and Costa Rica factory closure project, developing effective cost reduction plan and improvement strategies to project deadlines.