I build forecasting infrastructure for utilities and data centers.

Utilities forecast the electricity your home, your office and your city will need weeks ahead. Data centers have to forecast their own load the same way. When the forecast is wrong it costs money, and it is wrong often. The tools most operators have for this are spreadsheets, hand-tuned models and a data science hire they cannot make.

Rassket is the fix I wanted to exist. It connects to an operator's own data, builds a model tuned to that utility or data center, forecasts peak demand, and turns the forecast into something an operator can act on the same day. It explains its reasoning in plain language. The final call stays with the operator.

Rassket went through Founders Inc in San Francisco in the fall of 2025 and CO.LAB's Energy Innovation Accelerator in Chattanooga in 2026, and joined the NVIDIA Inception program in September 2026.

I built the entire production stack:

• 01

  AutoML pipeline

  Selects, trains and validates forecasting models on an operator's own data, with no data scientist in the loop.

• 02

  Energy feature library

  The domain-specific features that turn raw utility and data center data into signals a model can forecast from.

• 03

  Live utility data connectors

  Pull live operational data into the models as it arrives.

• 04

  Agentic operator layer

  Turns a forecast into a same-day decision, in the language an operator already uses.

I am a machine learning engineer and an economist. Both halves matter. Forecasting a grid is a statistics problem. Deciding what to do with the forecast, and what a wrong one costs, is an economics problem. With Rashmi Nagpal and David Guy Brizan I co-authored A Survey of Conversational Styles and Systems, published by IEEE at the 16th International Conference on Human System Interaction in 2024. Rassket is the second company I have founded.

I hold a double-major BSc in Mathematics and Economics and an MSc in Applied Economics with a focus in Data Science from the University of San Francisco. Since January 2024 I have also been a research assistant in the university's Computer Science Department, working on conversational style and on improving speech recognition systems with natural language processing and machine learning. In that work I kept running the same loop: clean the data, preprocess it, train a model, start over. The platforms that claimed to fix that were either black boxes or too shallow to trust. Rassket started as the tool I wanted for that loop, built for the one domain where I thought the error mattered most.

I am also Deputy Director of the West Africa Centre for Social and Economic Research, an independent non-profit research and policy institution based in The Gambia and the UK. My research there has focused on rural communities, disadvantaged communities and women farmers.

I am from The Gambia. I am the first college graduate in my family. Some villages there still have no power at all, and the grid that does exist is not something anyone takes for granted. I now build forecasting infrastructure for the American grid. I do not treat that as a story. I treat it as the reason the work has to be right.

The name is deliberate. Ras for Ras Tafari, Haile Selassie of Ethiopia, the one African country never colonised. Ket from rocket, because the whole point is speed.