Tyler Smith

I am an Applied Machine Learning and Artificial Intelligence Engineer focused on using traditional machine learning techniques and advanced transformer-based LLM’s to find signal in noisy, real-world data. I enjoy taking end-to-end ownership—from data wrangling to modeling, experimentation, and making results usable for business decisions.

At Apziva, I’ve applied traditional machine learning and statistical techniques to identify irrelevant survey questions, and I optimized a Stochastic Gradient Descent model to achieve a 95% customer satisfaction prediction. I also experimented with sampling techniques to reduce call center call-time by nearly 400 hours, and I identified a cohort of bank customers nearly 2x as likely as the dataset mean to subscribe to a term deposit product.

I build with modern AI workflows too: I integrated a Retrieval Augmented Generation (RAG) system to assist hiring managers in their vetting process, and I developed a custom bitcoin trading agentic workflow with nearly 75% success in suggesting portfolio structure. I’m especially drawn to agentic systems that can answer questions and support better decision-making.

Beyond my main role, I’ve worked on projects such as customer satisfaction analysis, call time reduction using clustering and large datasets, and an agentic bitcoin trading system using vectorized backtesting. My approach is iterative and detail-driven; clean data, test assumptions, and communicate findings clearly.