Benjamin Grayzel

Data Scientist & AI/ML Engineer

I build tools, forecasts, and models that make complex ideas clear and actionable.

About Me

I'm a data scientist and software engineer with experience building web apps, forecasts, and AI tools.

I built a 2024 US presidential election forecast in R, featuring simulation-based odds, interactive maps, and a sandbox playground. I've also worked with deep learning models and enjoy building side projects that explore politics, simulation, and forecasting.

In my spare time, I enjoy reading, climbing, automating fun projects, and writing on my substack. I'm currently pursuing my Master's in Computer Science at Dartmouth College (2026), where I graduated with a B.S. in Mathematics and Government (2024).

Skills & Expertise

  • AI/ML Engineering
  • Data Science
  • Forecasting & Modeling
  • Full-Stack Development
  • Python, R, C/C++, SQL, React

Experience

EXPLOR Intern

January 2024 - March 2024

Farallon Capital Management, San Francisco, CA

Developed & deployed full stack pipeline to visualize prediction market data from multiple sources, augmenting centralized search & info page. Data pulled from APIs on a daily cycle, merged/processed in unified schema, inserted into database, pulled dynamically and visualized in a search engine augmented with AI keyword extraction and semantic filtering.

Full StackVisualizationAIData Engineering

Financial Analyst Intern

June 2023 - December 2024

CND Life Sciences, Phoenix, AZ

Developed and deployed an interactive web application using R markdown's shiny framework to help sales team interface with customer data. Leveraged R to transform, analyze, and visualize customer data, enhancing data-driven decision-making for sales and leadership teams.

RShinyData AnalysisWeb Development

Featured Projects

2024 Election Forecast

A comprehensive presidential election forecast built from scratch with simulation-based odds, interactive maps, and data visualization. The model design begins with a weighted average of election polls, then uses backdated election data to estimate the distribution of various polling errors, simulating outcomes with a Monte Carlo methodology.

RShinyStatisticsData Visualization
View Project

Other Projects

Multi-Agent Framework Evaluation

Designed and evaluated an agentic framework using MCP, LangGraph, and GitHub for LLM-based code generation and tool-use benchmarking. Analyzed emergent behaviors in multi-agent systems through comparative evaluation strategies.

PythonAI/ML EngineeringMulti-Agent SystemsLangGraph
View Poster

Viral Information Research Model & Interface

Interactive model simulating the spread of viral true and false beliefs across a random network. Explore how propoganda propagates through social connections with customizable parameters.

RPythonNetwork AnalysisShiny
View Project

Physical Fluid Simulation

Advanced fluid simulation project based on the Lagrangian formulation of incompressible Navier-Stokes. Expanded to include 3D simulation, solid body coupling, irregular containment barriers, and interbody friction.

PhysicsSimulationMathematical Modeling

Eye Movement Classifier for Realtime Neurofeedback

Developed low latency REM and LRLR classification models for use in a real-time EEG-based system, integrating a complete neurofeedback loop with web-based monitoring and overnight deployment for lucidity induction.

PythonAI/ML EngineeringSignal Processing

LLM Toxicity Steering for Interpretability

Developed and evaluated an activation steering method for LLM toxicity control by training a linear probe on mid-layer activations in Gemma-series model, achieving significant manipulation of toxicity levels on in-distribution prompts.

PythonAI/ML EngineeringPyTorch

Substack

I write about politics, forecasting, and data science on my Substack. Here are some of my recent articles: