Senior Data Scientist — Predict Team

Remote, Eastern Europe or LatAm · Contract (B2B)

About the Company

We are a quantitative investment firm based in San Francisco. The firm automates investing decisions based on business fundamentals, combining data science, machine learning, and disciplined engineering practice, and runs a market-neutral long/short equity strategy driven by its own machine-learning platform.

The team is small and distributed across North America and Eastern Europe. Two of the firm's stated principles shape the research work directly: reason from first principles, and communicate openly and precisely.

Role Overview

The Predict team builds the models that decide what the firm buys and sells. As Senior Data Scientist, you will own research on models that forecast company-level performance from fundamental data, and your results feed the firm's automated strategy directly.

This is a research role with production discipline. You are expected to take a business question, turn it into a modelling problem, and defend the result — not just produce a notebook, but own the full path from question to production-ready signal.

The Problem, In Concrete Terms

Most quantitative roles are described in generic terms. This one isn't, so you can judge the fit before applying:

• Data: Tabular panel data with a time component — many features per row, built on fundamental company data. This is not classic single-series time-series forecasting, where you have one series and reconstruct its future.
• Universe: Roughly 2,000 US-listed companies.
• Target: Whether a given company's alpha will be good or bad over the coming period. The work is framed as regression on a defined target, not classification of entry and exit signals.
• Frequency: Medium — predictions are refreshed and the portfolio rebalanced on a daily-to-weekly cadence. This is not high-frequency trading.
• Models: Gradient boosting is the baseline. Neural networks have not reached the same level of performance on this data.
• Scale: On the order of one to two million rows. It runs comfortably on a single machine — there's no distributed processing to operate.
• Stack: Python and AWS.

Why Join

This is an opportunity to work directly with the people who set investment priorities at a small, capable, and motivated quant team — not as a downstream contributor, but as the person who owns the models the firm trades on. You'll work through the full research cycle yourself, present your findings weekly (including work in progress), and see your results feed straight into a live automated strategy.

Candidate Profile

The ideal candidate has extensive hands-on experience building predictive models on financial time-series data, is comfortable defending research decisions to a technical team, and understands the specific pitfalls of point-in-time financial datasets — leakage, survivorship bias, and validation schemes that respect time. You should be as comfortable maintaining reproducible research code as you are exploring a new feature set, and able to communicate partial or negative results as clearly as finished ones.

What you will do

• Design, develop and improve models that predict company-level performance from fundamental data.
• Take high-level business objectives set by the firm and turn them into research plans, then deliver against the priorities agreed with the team.
• Work through the full cycle yourself: understand the business problem, prepare the data, build and evaluate the model, and explain the result.
• Maintain research code so that results used in reports are reproducible by others on the team.
• Present progress and findings weekly, including partial findings on work that is not finished.
• Work with the data engineering side on the pipelines and features your models depend on.

Tech stack

• Python, Jupyter, NumPy, Pandas, scikit-learn
• CatBoost, XGBoost, LightGBM or equivalent gradient boosting packages
• SQL
• AWS, in particular S3 and Athena; SageMaker and Step Functions are used
• Git and GitHub, pull requests and code reviews, unit and integration tests, continuous integration
• POSIX-like environments: Linux or macOS, bash, ssh

Requirements

• Extensive education in mathematics, statistics, computational science, computer science or similar field (PhD is a plus but not required).
• 7+ years of professional hands-on data science experience.
• 3+ years working specifically on time-series prediction modeling. Experience on equity fundamental trading preferred.
• Explicit experience with financial time-series data machine learning pitfalls: point-in-time database construction, proper handling of delistings/survivorship bias, lookahead/leakage detection, and time-series-aware validation schemes.
• Hands-on familiarity with common software engineering practices and tools (git and github, pull requests and code reviews, unit and integration tests, continuous integration).
• Working knowledge of posix-like environments (Linux/macOS, bash, ssh) and AWS (Athena, S3, Sagemaker, State Functions) is strongly preferred.
• Excellent verbal and written communication in English.

Strong Pluses

Finance is a strong plus rather than a formal requirement. Beyond direct trading prediction on a next-day or next-week horizon, we also count:

• Price or volume prediction.
• Any task combining a time component with a large number of variables on a substantial data volume.
• A quant background, including portfolio construction and optimization work.

Nice to have:

• Machine learning applied to financial data, especially equities.
• Experience at a quant fund, proprietary trading firm, or systematic investment manager.
• Understanding of fundamental equity analysis and the data that supports it.

Terms

• Remote, from Eastern Europe or LatAm.
• Daily overlap with the US team from 6:00–9:00 AM Pacific Time (15:00–18:00 CEST / 16:00–19:00 EEST).
• Compensation from USD 7,000/month, and above that for exceptional candidates.
• Performance bonus of 25–50%, based on firm performance and individual performance.
• B2B contract
• Daily stand-ups and a weekly review, where you present progress and findings.
• The firm is not able to contract with candidates residing in Russia.

Hiring Process

1. CV review
2. Intro call with Recrucial
3. Technical screening call with a data scientist from the firm
4. Test problem on real data
5. Interview on the test problem
6. Final interview with firm leadership (1–3 meetings)