Built an automated system for large-scale influencer data collection, processing, and monitoring using AWS servicesDeveloped and executed an automated data collection pipeline using YouTube API and Selenium to retrieve influencer data from over 10,000 YouTube channels across multiple industriesImplemented AWS Batch processing to efficiently handle the large-scale insertion of data into DynamoDB, enabling parallel processing of thousands of records to optimize performance and ensure scalability for future data expansionSet up comprehensive monitoring and logging for data collection and processing workflows using AWS CloudWatch, ensuring real-time tracking of job status, DynamoDB performance, and system health

Developed Yelp recommendation systems based on user rating predictionsImplemented an item-based collaborative filtering recommender system using Pearson similarity; leveraged Python and Spark RDDs to handle datasets with 1M rows to predict Yelp user’s preferences and enhance personalized recommendationDeveloped a machine-learning-based prediction system with XGBoost and feature selection, achieving RMSE 0.87Integrated similarity-based and machine-learning-based recommender with an adaptive weight function, reducing RMSE to 0.79 on out-of-sample data, showcased real-world applicability in personalized business recommendations

Established a fine-tuned rating model for enterprises worldwide and maintained a comprehensive client databaseDeveloped a regression model summarizing enterprise financial ratings based on discretized output of logistic regression and random forest that are directly included in the final rating along with manual review and other documentsReconciled and rectified USCid (Unified Social Credit Identifier) in company database with official commercial registry using web scrapingMigrated over 100k rows of client-product data from local files to MySQL database and integrated with new CRM system, improving sales operations efficiency by 20%Analyzed discrepancies between system data and credit rating reports generated over the past year, focusing on the validity period of the ratings and created reports and visualizations using Tableau and presented the identified patterns to analyst team

Led a team to implement a distributed MySQL database system on EC2 for university dataEnhanced database scalability by dynamically creating and distributing new databases based on increased demand, utilizing a hash function to allocate students across different databases efficientlyDeveloped a backend system that manages data, integrating advanced preprocessing and validation scripts to ensure the integrity of data entries, with front-end capabilities for robust administrative functions including record retrieval and updatesConfigured MySQL databases, defined schemas for key entities, and populated databases with detailed data entries to ensure functional integrity and relational data linkage

Teaching students 1-on-1 for various subjects, including Data Science, statistics, programming, Calculus courses, and GRE, SAT, and TOEFL test preparation.

Facilitated the coordination of key meetings and rigorously evaluated portfolios to identify top talent. Expertly assessed game user experiences, providing detailed reports to support product development and refinement.

Built a predictive model that leveraged historical stock data for future stock pricing using a Recurrent Neural NetworkVisualized trend and seasonality of stock log return time series data with EDA and developed SARIMA prediction model including ADF test to confirm stationarityDeveloped a stock price prediction model using PyTorch-implemented LSTM neural networks on 10 years of market data, focusing on feature enhancement through MinMax scaling for data preprocessingReduced model's loss to 0.07 over 300 epochs using Adam optimizer coupled with Mean Squared Error (MSE) loss criterion; optimized the model with hyperparameter tuningValidated the model through data visualizations, demonstrating tight alignment with actual market trends, which garnered stakeholder approval and was subsequently adopted for financial analysis and decision-making