Dylan Mcintosh Email & Phone Number

Orange, CA, US

Orange, CA, US

• Utilized cloud computing to host and develop a local in-house LLM API in Python to cut costs from dependence on OpenAI API calls
• Stress tested the Linux server to ensure that it fits the needs of the company’s API call load during peak hours
• Analyzed and tested several open source HuggingFace LLMs for a low-latency, low-cost solution on different CUDA hardware in the cloud
• Determined Mistral 7B running through the vLLM inference engine to be the most cost-effective framework
• Developed RAG software to scrape internet for relevant information and drastically improve product output quality

• Analyze state test scores & cognitive metrics in youth to determine relationships of mental health, cognitive metrics, & school performance
• Cleaned the underlying data & used best practices in Python to revive an abandoned project with no documentation & no code
• Trained linear regression models to identify promising features for student performance prediction & inform the project roadmap
• Utilizing random forests to rank feature importance & determine which cognitive metrics are the most valuable to analyze further

Remote

• Analyzed the stepwise rate segmentation model to measure risk & price premiums to find what to improve upon
• Developed a functional data pipeline framework to test many different variations of models in Python with pandas, sklearn, & H2O
• Employed software engineering best practices with thorough documentation in docstrings to help R&D team transition from SAS to Python
• Developed custom metrics in the form of relative MAE & Lorenz curves/GINI coefficient to compare model performance
• Created new GLM models that improve upon the baseline household demographic model of the current rating structure
• Developed a web application using ReactPy to improve the timesheet process, became implemented companywide by HR to cut costs

Orange, California, United States

• Analyzed VR positional and rotational data with over 2 million rows to develop models that understand human-VR interaction using Python
• Collaborated with team members on organizing tasks and maintaining version control via Git
• Conducted exploratory data analysis using plotnine to visualize data, found we need to adjust positional data from relative to absolute
• Optimized algorithm efficiency by about 60% to maintain sufficient runtime over 2 million rows of data
• Engineered a series of novel features to serve as inputs to machine learning models to validate hypotheses on the underlying data
• Currently planning which models to utilize to determine actions subject is taking (reaching, tossing, moving) using said engineered features

Remote

• Developed Python code to find indirect profitability of digital products by determining which products brought valuable returning customers
• Visualized the calculated profitability using plotnine to make the data interpretable
• Interpreted and presented the output to the business owner to assist in making critical business decisions about future marketing strategies.
• Ingested and cleaned over 200 GB of tabular data into pandas dataframe then exported into a SQL database for company usedecisions.

Master Of Science - Ms, Electrical Engineering And Computer Science

Bachelor Of Science - Bs, Data Science