Quantitative Developer - Agentic Research Lab

About Flowdesk

At Flowdesk, we are building the infrastructure powering the next era of global finance. As a full-service digital asset trading and technology firm, we provide liquidity solutions, OTC trading, and market infrastructure that connect digital assets to global markets with institutional-grade precision, transparency, and scale. Flowdesk combines expertise from traditional finance, algorithmic trading, and blockchain technology to deliver always-on trading capabilities across the digital asset ecosystem.

Founded in 2020, we have grown into a global team supporting institutions, exchanges, and token issuers worldwide as the financial system evolves toward tokenized markets.

You’ll be joining a company operating at the intersection of finance, technology, and crypto - solving complex market problems, building critical infrastructure, and helping shape how digital assets move globally. We move fast, think long term, and work with a strong ownership mindset across engineering, trading, product, operations, and business teams.

Research infra is a new team inside Technology, turning quantitative research into production by owning the research platform: the market-data store, the simulation framework, and the compute that lets quant researchers test thousands of hypotheses a day. Tick-level order-book data and our own execution traces stream through Redpanda into Delta Lake and ClickHouse. Two Quantitative Developer tracks make that platform fast (Research systems) and usable (Research tooling); this role adds the analytics that keep running when nobody is at the keyboard.

Mission

You’ll design and lead the agentic analytics of the platform. The first goal is to automate post-trade analysis (PTA): agents analyse every live strategy continuously, compare it with simulation, triage anomalies and write up what they find, while researchers review the conclusions. Once that loop is trusted, the agents move upstream to hypotheses and experiments. You’ll set the architecture, ship the first production agents yourself, and define the guardrails that keep automated analysis honest. You’ll work closely with quant researchers, traders and the risk team, as well as the Research systems and Research tooling engineers whose platform your agents run on.

Tasks

Automate post-trade analysis

• Build agents that produce PTA for every live strategy leg on a schedule: P&L attribution, markouts, adverse selection, fill rate, inventory and fees.
• Replace the notebooks and dashboards researchers maintain by hand today.

Compare simulation with reality

• Replay live periods through the simulator and compare simulated fills and P&L with what actually happened, per strategy leg.
• Flag drift before it costs a week of P&L.

Build monitoring with agentic triage

• Turn PTA deltas, sim-vs-live drift, P&L, inventory and exposure into alerts.
• Have agents triage them first, and hand a summary and recommended actions to the trader on call.

Plug agents into live data

• Consume market data and execution logs from Kafka / Redpanda and query them in ClickHouse, so monitoring and research agents read data that is seconds old, not T+1.
• Expose the platform as typed tools - for example an MCP server over the metrics library, data access and the simulator - so agents call the same validated code paths researchers do.

Make it production-grade

• Version, test and deploy reports, datasets, prompts and agents through CI/CD (GitHub Actions, GitOps with Flux / Argo CD).
• Land every agent output with its config, data snapshot and code version, so nothing it produces is unreproducible.

Evaluate the agents themselves

• Build eval suites with known-answer tasks (planted anomalies, look-ahead traps) and track how often agents find, miss or fake them.
• A regression in the eval suite fails the PR.

Later: from findings to hypotheses

• Extend the loop upstream: agents draft hypotheses and campaign specs, run them on the simulation cluster, and a critic agent tries to break the result before a human sees it.
• Build in the defences: trial budgets, multiple-testing corrections, untouched holdouts agents cannot read, and independent re-execution of every claimed result.

Ways of working

• Ship through code review and CI like the rest of engineering - no exceptions for agent-authored code or prompts.
• Document architecture decisions and the guardrails that keep automated analysis honest, so another engineer can operate the loop without you.
• Own your systems end to end, from design through production monitoring and incident response.

How we’ll measure your impact

• Researcher hours per week spent on PTA (down), and the share of live strategy legs with PTA that refreshes on its own.
• Time from a live anomaly to a triaged summary in front of a trader.
• Share of agent-reported findings that reproduce on independent re-execution.

Requirements

Must Have

• 10+ years building production software or ML systems, including several years as a technical lead who set architecture for other engineers.
• AI agents in production, not demos: multi-agent orchestration, tool calling, evaluation and red-teaming, cost and model routing, staged rollouts with a human in the loop.
• Hands-on quantitative experience: you have built backtests, factor or signal models, or risk models, and have a healthy distrust of results that look too good - look-ahead bias, data snooping and regime change are second nature.
• MLOps and CI/CD at platform level: you have built (not just used) model registries, experiment tracking, automated validation and deployment pipelines, and treat reproducibility as a hard requirement.
• Streaming and analytical data systems in production: Kafka (or Redpanda) and ClickHouse, including schema design, partitioning, consumer lag and query performance at high volume.
• Strong Python and SQL.
• At ease on Linux and Kubernetes, with infrastructure-as-code for deploying and operating workloads.
• Judgement to know which parts of research to automate, and which must stay with a human.

Nice to Have

• Production experience in a systems language - Rust, C++ or Go - with Rust especially valuable, since the trading engine and simulator are written in it.
• Experience in a trading environment, ideally market making, HFT or crypto.
• Market microstructure knowledge: order books, queue position, markouts, adverse selection.
• Open-weight model serving (vLLM, SGLang or similar) on your own GPUs.

Benefits

• International environment (English is the main language)
• Team events and offsites
• Learning and Development budget
• Health insurance for employee and legal dependents

Recruitment Process

Our process is consistent across all roles and is designed to assess genuine technical depth. Candidates should expect the following stages:

• Stage 1 - Initial screening with TA (30 mins)
• Stage 2 - Technical Screen - Meet the Data Director (45 mins)
• Stage 3 - Quant Interview - Meet the Quant Developers and Quant Researchers teams (60 mins)
• Stage 4 - Roundup Interviews - Meet the Infrastructure Director, the Data team, and the Chief Technology Officer (90 mins)

On the agenda, discussions rather than trick questions! These moments of exchange will allow you to understand how Flowdesk works and its values. But they are also (and above all) an opportunity for you to present your career path and your expectations for your next job!

We are committed to an inclusive and accessible recruitment process. If you require any reasonable adjustments or have specific needs to enable you to participate fully in the interview or assessment process (e.g., a sign language interpreter, extra time for a test, or an accessible location), please contact us to discuss how we can support you.