The Forecasting Company
New foundation models for time series
The Forecasting Company builds planning systems based on our in-house foundation models for time series. We are starting with forecasting problems for logistics.
Active Founders
Joachim Fainberg, Founder & CTO
PhD in ML for speech and full-stack developer. Interned at Sonos where I built an on-device wake word detector and at Bloomberg building a large ASR system. Worked full-time at JPMorgan in their Machine Learning Center of Excellence in New York, and then at two startups, team lead ML/NLP at Consigli, a prop-tech startup, and sr algorithm engineer and full-stack at Vind AI, a SaaS for designing off-shore wind parks.
Geoffrey Negiar, Founder & CEO
Co-founder and CEO at The Forecasting Company, helping enterprise customers improve their operational decision making. PhD grad from UC Berkeley's BAIR. I spent time with Amazon Forecasting Science in SCOT, Google Brain, and Bloomberg LP where I worked on forecasting, optimization and deep learning research.
Company Launches
tl;dr: The same core technology that led to the LLM explosion can be adapted to build accurate and easy to use time series models. We’re building the enterprise-grade, plug and play forecasting systems of tomorrow. Talk to us if you have forecasting and planning use-cases!
Hi everyone, we’re Geoff and Joachim. We’ve built ML and forecasting systems at Amazon, Google, Bloomberg LP, JP Morgan, and Sonos. We’re The Forecasting Company!
Problem 📉
- Existing time-series forecasting solutions produce unreliable predictions with sometimes catastrophic results - they have a hard time handling activity spikes like Black Friday.
- Developing and maintaining useful forecasting and planning models requires large and costly teams of data scientists and domain experts.
- Data science teams are backlogged, leaving operators waiting for months before a use-case to be put into production.
Solution 📈
- An easy to use, flexible and accurate forecasting system. No training required!
- A massive training dataset of time-stamped data from many industries and domains (energy, transportation, etc.).
- A forecasting model that can leverage user-provided input variables or our own curated data streams.
What we’re forecasting these days
Bacteria levels in the Seine 🦠 🏊♂️
Our forecasts show that swimmers lucked out today 🍀!
(Data provided courtesy of the Fluidion Open Data Initiative)
Ask
- If you have forecasting problems in-house, or know of someone who does, please reach out - we’d love to help!
- Try our sandbox model on www.theforecastingcompany.com and let us know what you think
founders@theforecastingcompany.com