(Junior) Quantitative Researcher / Developer - Systematic Equity (m/f/d)

Über die Position

Ultramarin is a quantitative asset manager in Berlin. We run systematic equity (long‑short and long‑only) and asset‑allocation strategies in developed markets.

You’ll join the Equity Selection team, which builds machine‑learning models to forecast relative stock returns.

Your focus will be the alpha signals that power these models — work that combines research, economic intuition, and engineering.

We care not only about finding signals that work, but also about understanding why they work.

This is hands‑on work: what you build drives our live strategies.

Deine Aufgaben

You’ll start by building and evaluating individual signals, and your scope will grow as you do.

• Search for and evaluate signals: Look across datasets and investment universes for signals that hold up out of sample, and assess both statistical and economic performance.
• Implement alpha signals: Transform raw, noisy data — prices, trading volumes, fundamentals, text — into robust alpha signals.

Build them as small, tested, parameterised nodes in our feature computation graph, while avoiding look‑ahead bias.

• Construct composite signals: Combine many correlated signals into a few robust composites per universe, weighting them by uniqueness and information content, and separating genuine stock selection from unintended static allocation tilts.
• Automate the research loop: Build and use LLM/agentic tooling (e. g. in Claude Code) that discovers signals, runs experiment grids, and generates evaluation reports.
• Dein Skillset

To thrive in this role, you should bring

• A master’s degree or Ph D in mathematics, physics, computer science, financial engineering, statistics, or a closely related quantitative field.
• A solid grounding in statistics and a working knowledge of econometrics, with the ability to reason about noisy real‑world data.
• Familiarity with collaborative Python development (Git, code review).
• The ability to write maintainable, well‑tested code (we use Pydantic and pytest).
• Experience manipulating data with Polars or pandas.
• A strong interest in financial markets and the quantitative investment process.
• A team‑oriented mindset and a preference for in‑office collaboration

Nice to have

• Experience identifying and evaluating alpha signals for systematic or fundamental equity strategies.
• Experience with ML libraries such as scikit‑learn, Light GBM, or Py Torch.
• Experience building LLM/agentic tooling and evaluation.
• Experience with numerical, statistical, and MLOps libraries such as Sci Py, statsmodels, and MLflow.

Warum wir?

• A research environment where ideas move quickly from prototype to production.
• Mentorship from senior researchers and engineers, with early ownership of your work.
• A collaborative, interdisciplinary team spanning quantitative finance, software engineering, and machine learning, with a shared focus on quality, openness, and attention to detail.
• Internal workshops and team events, Urban Sports Club membership, and access to Corporate Benefits.
• A loft office in Berlin (Prenzlauer Berg), healthy food and drinks, and Apple hardware.
• Über uns

Wir sind ein Deep-Tech-Pionier und bieten unseren Kunden KI-basierte Investmentlösungen.

Aufbauend auf den Best Practices der quantitativen Vermögensverwaltung erschließen wir das Potenzial des maschinellen Lernens für nachhaltige Investitionen am Kapitalmarkt.

Unser interdisziplinäres Team besteht aus Experten in den Bereichen Finance, Informatik, Software Engineering, Machine Learning sowie Mathematik, Physik und Neurowissenschaften.

Ultramarin ist durch einen engen Austausch mit führenden Universitäten und als Mitglied von Inquire Europe in die globale KI-Community eingebunden.

Ultramarin ist einer der Vorreiter für KI-basierte Analysen und Entscheidungsprozesse im Asset Management.

Seit der Gründung in 2017 hat Ultramarin seinen Hauptsitz in Berlin, sowie weitere Standorte in Frankfurt und München und wird von führenden internationalen Business Angels und VCs unterstützt.

#J-18808-Ljbffr