Quantitative Developer – Agentic Research Lab

About the Role
In this role, you will serve as the technical lead for the agentic research lab, setting the architecture, shipping the first production agents, and defining the guardrails that ensure the integrity of an automated research loop. You'll collaborate closely with quant researchers, traders, and the risk team, as well as with Systems and Research-tooling developers, ensuring that your agents can operate effectively within the research platform.

Mission
Your mission includes leading the transition of the research platform to production-grade standards, ensuring the agents can seamlessly integrate into live data streams, and designing a robust multi-agent workflow that drives continuous improvement in research processes. You will oversee controlled experiments in production and construct measures to prevent overfitting in research models.

Key Responsibilities

• Make the research platform production-grade (MLOps & CI/CD)

  • Treat research like software: version, build, test, and deploy every strategy, signal, dataset snapshot, and agent prompt through CI/CD processes.
  • Own the model lifecycle end to end, ensuring a traceable link between model performance and its code, data, and verdicts.
  • Ensure agents follow the same rigorous standards for tools, prompts, and evaluations through CI, preventing regressions.

• Plug agents into the live data streams

  • Consume market data and execution traces from streaming services such as Kafka or Redpanda, ensuring availability to research agents in real-time.
  • Design data topics, schemas, and views for effective monitoring and research capabilities.

• Build the research agent loop

  • Develop a multi-agent workflow that includes hypotheses generation, experiment design, execution, and analysis.
  • Utilize typed tools to facilitate agent interaction with validated code paths for consistency in research outputs.

• Make “continuous” real

  • Run scheduled research rounds to re-evaluate live strategies and suggest improvements, ensuring all results are registrable for reproducibility.

• Build continuous monitoring with agentic triage

  • Create alerts based on performance indicators and have agents initiate triage, summarizing incidents for trader review.

• Run controlled experiments in production

  • Manage the phases from shadow mode to live trading while implementing necessary safeguards, and establishing criteria for evaluation and progression.

• Guard against automated overfitting

  • Implement restrictions on trials per hypothesis, and corrections in multi-testing to maintain research integrity.

• Evaluate the agents themselves

  • Develop evaluation suites to continuously assess agents' capabilities and improve their performance based on known metrics.

• Run models on our own hardware

  • Establish secure on-prem GPU inference and define tasks for external API usage to safeguard proprietary data.

Ways of working

• Collaborate through code review processes, ensuring thorough documentation of architectural decisions and monitoring systems to facilitate teamwork and operational transparency.
• Take ownership of your systems' design, production monitoring, and incident response.

How We’ll Measure Your Impact

• Number of research rounds completed per week without human intervention and the percentage that culminate in a reviewed verdict.
• Proportion of agent-reported results that confirm during independent re-execution.
• Response time from anomaly detection to trader alerts.
• Successful promotions of strategies from shadow to live, with an aim to minimize incidents along this path.

Requirements
Must Have

• Over 10 years of experience in building production software or ML systems, with significant experience as a technical lead.
• Practical experience with multi-agent orchestration and automated decision systems in risk-sensitive environments.
• Hands-on quantitative background in developing backtests, models, and risk assessments, familiar with biases and market complexities.
• Expertise in MLOps and CI/CD, including model registry and automated pipelines with reproducibility as a priority.
• Experience with streaming and analytical systems like Kafka/Redpanda and ClickHouse, with a strong grasp of performance optimization.
• Proficiency in Python and SQL, with familiarity in Linux environments and Kubernetes.
• A solid understanding of regulatory and compliance frameworks regarding automated trading systems.

Nice to Have

• Experience with systems programming in Rust, C++, or Go, with a preference for Rust due to its use in trading engines.
• Familiarity with trading environments, particularly in market making, HFT, or crypto trading.

Benefits

• Work in an international environment with English as the primary language.
• Participate in team events and offsites.
• Access to a learning and development budget.
• Health insurance coverage for employees and legal dependents.