Most retail investors operate without the quantitative tools that institutions take for granted. We are building those tools — calibrated, transparent, and accountable to a public record. The work is technically difficult, intellectually demanding, and consequential.
A Bloomberg Terminal costs $24,000 a year. Inside hedge funds and banks, analysts have direct access to calibrated risk models and probabilistic forecasts. Outside those institutions, the same questions are answered with intuition and confident guesswork.
ATLAS exists to close that gap. We run Monte Carlo simulations on every US-listed stock, calibrate the outputs against a public ledger of 285,907 timestamped predictions, and deliver the result as honest probabilities — not opinions, ratings, or signals. Current accuracy: 95.4%.
These are not active openings. They reflect the disciplines we expect to recruit for as the company scales.
Core simulation engine, regime classification, calibration measurement. Strong foundations in statistics and probability.
Consumer product surfaces with clear ownership of latency, reliability, and end-user experience.
Probabilistic interfaces designed for clarity rather than engagement.
Distribution to retail investors at scale, without compromising on product principles.
Remote with strong overlap to European hours. Periodic in-person work in Madrid.
Competitive base salary plus meaningful equity. Discussed during the process.
Two structured conversations followed by a paid evaluation project.
Strong written communication, quantitative reasoning, serious orientation toward the people we serve.
A model well-calibrated on 100 predictions can drift across 100,000. Maintaining 95.4% accuracy across 4,637 stocks demands constant attention to data integrity, regime detection, and edge cases.
Describing Monte Carlo in a textbook is straightforward. Keeping the engine running through earnings, macro shocks, and structural market changes — and reporting honestly when the model is uncertain — is harder.
Most retail growth is built on engagement loops and gamified UI. We have decided not to use any of that. Reaching the audience anyway is a serious strategic challenge.
Probability claims only become credible across thousands of forward-looking observations. The discipline required to build a brand on accuracy compounds slowly and breaks quickly.
Five principles that define how the team operates.
New hires take responsibility for full product surfaces, not isolated tasks. The best work happens when one person ships a feature end-to-end and owns the consequences.
Decisions are tied to measurable outcomes — calibration accuracy, latency budgets, prediction accountability. Claims are validated before they are published.
No real-time on-call rotations. Speed comes from clarity, not urgency.
The 285,907-prediction ledger is public. Methodology is documented. Internal communication mirrors that standard.
We optimize for compounding accuracy over the next decade, not short-term feature velocity.
The team is small. Whoever joins next will own real surface area — whole product lines, whole layers of the engine, whole motions of the company. There are no large teams to hide behind.
We hire generalists who can move across product, engineering, and quantitative work, and specialists who want to spend years pushing one discipline forward. Both shapes of person are essential.
Submit your details below and attach a CV. Submissions are sent directly to info@atlas-stocks.com. We review every introduction and respond when there is a relevant fit.
Read our principles and methodology.