About
Why we built it
Dahlia spent six years building risk and forecasting models at a consumer bank. She watched the data predict, with two or three days of lead time, that an account was about to go negative. Then she watched the fee land anyway. The bank had the information. The customer did not.
That is not unusual. Banks earn significant revenue from overdraft fees, and the infrastructure to project a customer's cashflow has existed for years. What has not existed is a product that hands that projection back to the person it concerns, in a form they can act on before it matters.
Arjun had spent a decade building sequence and time-series models at scale. The question that started the company was simple: how hard is it to build a forecast that actually tells someone their tight week is coming? Not after. Before.
Forecasting an individual's cashflow is a real machine-learning problem. Recurring-pattern detection, confidence estimation, anomaly flags, and the ability to stay honest when the income is irregular. Tellwise pairs a predictive model over your transaction history with a language model that explains the forecast. The numbers come from code. The model writes the read.
The team
Five of them are below. Everyone is based in the United States. Tellwise is what we do.
Co-founder and CEO
San Francisco, CA
Spent six years building risk and forecasting models at a consumer bank, where she watched people get hit with overdraft fees the data had seen coming days earlier. Tellwise is the warning the bank never sent them.
Co-founder and CTO
San Francisco, CA
Ten years in applied machine learning on time-series and sequence models, most recently forecasting demand across millions of accounts. He owns the predictive engine, the pattern detection and the evaluation set every change is measured against.
Head of Data Science
San Francisco, CA
Builds the models that separate a recurring subscription from a one-off charge, and a normal week from a spike. Her rule: never show a warning we cannot explain in one plain sentence.
Founding Engineer
Austin, TX
Works on the parts nobody sees: parsing a bank export that four different banks each format differently, and keeping a forecast honest when the history is thin. He owns the ingestion pipeline.
Design
Brooklyn, NY
Designs the forecast screen, which is the whole product. Her rule: someone should know whether next week is tight within four seconds of the page loading, without reading a legend.
The people and investors named on this page are illustrative. Tellwise is a Pre-Seed company and we would rather say so plainly than have you find out later. The product is real: the forecast engine, pattern detection and the free tier work as described.
How we work
A narrow confident band that turns out wrong costs someone more than a wider one that turns out right. We show the range and say plainly when the range is wide, especially when the history is thin.
If we cannot explain a tight-week flag in a sentence a person can act on, it is not ready to ship. Every warning carries its reason, in plain language, with the date it bites.
Tellwise shows what your own transactions suggest is coming. It does not tell you what to do with that. We are careful about where that line is, because it matters.
Nine people, none of us here part time. You own an area, talk to users in your first two weeks, and see what you built reach people before the next one does.
The company
Backed by
We raised $900,000 at Pre-Seed in November 2023, led by Foreshore Ventures. The funding goes into the product, not advertising it.
Foreshore Ventures
Lead investor, Pre-Seed
An early-stage fund that backs consumer software with a real data moat.
Nadia Halloran
Angel
Former head of product at a neobank.
Peter Osei
Angel
Founded and sold a personal budgeting app.