Rob Churchill — Machine Learning & NLP Expert

About my work

I am an artificial intelligence (AI) and machine learning (ML) engineer. In December 2021, I finished my PhD in Computer Science from Georgetown, where my research modernized topic models to handle the noise and sparsity of real social media text. That work became the topic-noise model family, published at the ACM Web Conference (WWW), PAKDD, and Knowledge and Information Systems (KAIS), alongside a comprehensive survey of the field, "The Evolution of Topic Modeling," in ACM Computing Surveys.

I run Churchill Software, a small senior consultancy that takes clients from prototype to production, with recent work spanning an AI agent for automating construction takeoffs, and AI agent pipelines to automate financial analysis tasks, as well as AI advisement for fledgling startups. I work with all large language models (and the smaller ones as well!), and have deep experience with databases ranging from SQL to unstructured document stores. I spend a lot of time helping startups bridge the gap between a demo or proof of concept, and a production system that works at scale.

Experience

Head of Data Science

Current

5-Out2022-10-01 – Present

Direct predictive analytics and machine learning model development to improve restaurant performance. Design and implement the forecasting models, monitoring systems, and evaluation methodology behind key restaurant KPIs, and fold in new data sources (events, reservations, inventory) to give operators more accurate forecasts. Remote.

Owner

Churchill Software, LLC2022-01-01

Consultant to numerous small startups: I design, build, and advise on applications ranging from multi-agent AI systems to databases and application backends. As an advisor, I help startups plan, design, and direct the implementation of AI-first applications, database architectures, and cloud environments. As an individual contributor, I personally implemented simple and complex AI agents, production databases and APIs, and serverless cloud infrastructure. Remote, United States.

Data and Machine Learning Consultant

Incentivio2024-02-01 – 2025-12-01

Designed and led a team to build cost-efficient database architecture, models, and infrastructure allowing the company to scale to much larger clients and to serve reliable data to APIs. Implemented machine learning and statistical models to classify customer data, and designed data visualizations and reports providing transparency into new and existing models. Remote.

Notable work

Topic-Noise Models (TND / GTM)

A family of topic models that explicitly model a noise distribution alongside topics, built for short, noisy social media text. Open sourced as GU-DataLab/topic-noise-models-source and the gdtm Python toolkit.

5-Out Restaurant Forecasting

Production forecasting and ML stack predicting sales, labor, and demand for restaurant operators, plus the evaluation methodology used to judge model accuracy.