Kevin Tran Email & Phone Number

Fremont, Ca, Us

Palo Alto, California, United States

• Collaborate with faculty and researchers to prepare and analyze large-scale datasets, offering consultation on research design and coding, while leveraging expertise in machine learning, text processing, data engineering, and high-performance computing (HPC) to advance Stanford GSB’s research objectives.• Notable projects include: o Designed a dynamic neural network with hyperparameter tuning using Ray Tune to identify the top 10% of individual stocks with the highest monthly return rates. o Developed and trained machine learning models on high-performance computing (HPC) servers with multiple GPUs, including: - Predicting photo manipulation levels in images. - Classifying patents for green technology usage. o Optimized large matrix multiplication code, reducing computation time from 6,000 seconds to 2 seconds. o Built data engineering pipelines to publish large-scale, high-security datasets to the Redivis platform. o Collected and analyzed diverse datasets (e.g., tweets, demographic data, SEC logs, and filings) using prebuilt APIs and custom web scrapers developed with BeautifulSoup to support research efforts.• Educated researchers on HPC best practices, including Python virtual environments, GIT version control, Slurm job submissions, and parallel computing optimization.• Conducted demonstrations on fine-tuning GPT-3.5, building models with the HuggingFace platform, and running large language models locally.

Redwood City, California, United States

• Served as a Technical Lead at Apple, recommending and designing a new Data Lakehouse infrastructure.o Implemented a proof of concept to efficiently develop AWS Airflow pipelines with Docker containers.o Led the team in establishing new standards for data ingestion, data storage, database schema, data pipeline development, Python coding, code version control, CI/CD, and Tableau dashboard.• Worked as a Senior Data Engineer at Apple, maintaining, enhancing, and implementing various features for the Airflow data pipelines in a complex code base for near real-time Business Monitoring.o Developed a Tableau dashboard to automate manual work, providing instant access to insights.• Functioned as a Senior Data Engineer at Meta, developing and enhancing multiple Dataswarm pipelines to automate manual accounting auditing tasks across all financial transactions.o Led and created a best practices guide for data engineering to facilitate onboarding for other Slalomers at Meta.• Functioned as a Senior Data Scientist at Meta, developing and deploying a NLP text classification model to production with FBLearner Flow and Fluent2 to flag harmful post contents.• Worked as a Senior Data Scientist at Charles River Lab, building an end-to-end ML server in Python and FastAPI to detect experiment anomalies.o Developed high-signal aggregated features for detecting erratic signals, spikes, dips, and goodness of fit in linear trends.o Wrote comprehensive documentation covering model development, API creation, and model deployment.• As a Senior Data Scientist at Slalom, developing a new Inclusive Recruiting capacity offering.o Built an API with four endpoints and conducted a demo in AWS.o Acted as an admin in Bitbucket, creating a git branching strategy, reviewing codes, and supporting other developers.o Established coding standards and improved the code base for consistency and maintainability.

Redwood City, California, United States

• Led the creation of standards for file repository, code version control, and project workflow for the internal team.• Developed an NLP model to classify spending in 22 categories for millions of transactions.• Engineered an NLP model to automate the identification of leases, saving accountants hundreds of hours.• Programmed a simulation to optimize staffing levels for a call center, minimizing missed calls and downtime.• Conducted network analysis to effectively measure the Wu Tsai Institute's impact on professors' collaboration.• Visualized the relationship between hundreds of authority functions (access rights) and their corresponding job titles.• Conducted statistical tests on cycle times to identify bottleneck factors for process improvement.• Created custom Tableau dashboards for the Merchant Services and Cash Management Group.

Mountain View, California, United States

First founding data scientist on the Robo-Advisor team spearheaded multiple data science initiatives:• Pioneered data science initiatives as the first data scientist on the Robo-Advisor team.• Implemented and deployed a propensity model, including A/B testing for validation. The model identified the top 20% of users with the highest product interest scores, accounting for over 50% of total revenue.• Developed a borrowing power model using a decision tree in Python and deployed it in Java.• Identified and addressed critical data issues, collaborating directly with the engineering team to enhance data quality.• Played a key role in recruiting, interviewing, and referring candidates to build a robust data science team.

San Diego, California, United States

LoanHero was a small fintech startup that provided instant web-based financing at the point-of-sale for merchants. As the founding data scientist, I worked directly with Chief Risk Officer on risk management and fraud detection accounting for over $100 million in total loan funding through the following projects. It was then acquired by LendingPoint in January 2018.• Spearheaded the data ingestion/ETL process, as our loan servicing partner lacked a data API. • Implemented a Python program to parse and clean over 50,000 HTML files, ensuring our database was consistently updated. This database served as the foundation for all analytics, reports, dashboards, and filings.• Conducted data exploration and feature engineering by parsing XML credit reports in Ruby to refine our credit model. Through ongoing experimentation, identified the most informative signals for our credit risk model and formalized it as the 'merchant risk score'.• Leveraged insights from defaults and the collections of features to define a 'credit-based loan limit'. This rule-set was analyzed in Python and deployed using Ruby in the production environment, resulting in significant savings for LoadHero by minimizing delinquent loan losses.• Forecasted charge-off rate projections by analyzing internal and external Lending Club data. This analysis contributed to informed decision-making regarding risk management strategies.

Master’S Degree, Engineering Management, 3.74

Bachelor’S Degree, Mathematics And Accounting, 3.48