Quantitative Developer – Agentic Research Lab

As a Quantitative Developer leading Flowdesk's Agentic Research Lab, you'll architect and ship the first production multi-agent research loop that formulates hypotheses, writes and runs experiments, validates results, and monitors live strategies. You work closely with quant researchers, traders, risk and platform teams to build a system where agents operate at the speed of market data, with humans reviewing verdicts, not boilerplate.

What you'll do

• Make the research platform production-grade by treating research like software: version, build, test and deploy every strategy, signal, dataset snapshot and agent prompt through CI/CD (GitHub Actions, GitOps with Flux / Argo CD), with model lifecycle owned end to end from experiment tracking through automated validation gates to staged deployment and rollback
• Consume market data and execution traces from Kafka / Redpanda topics and query them at scale in ClickHouse, so monitoring and research agents see fresh data within seconds, and design the topics, schemas and materialized views with backpressure, replay and exactly-once semantics where they matter
• Design a multi-agent workflow with a hypothesis generator, an experiment writer that turns hypotheses into campaign specs, an executor that runs them on the simulation cluster, and a critic that validates results before humans see them
• Expose the research platform to agents as typed tools via an MCP server over the research SDK, campaign engine and experiment registry, so agents call the same validated code paths researchers do
• Run scheduled research rounds on fresh data using Dagster to re-validate live strategies, re-run parked ideas when the simulator improves, and propose parameter updates, landing every round in the experiment registry with config hash, data snapshot and code version
• Turn sim-vs-live comparisons, PnL, inventory, exposure, latency and regime indicators into alerts, and have agents triage alerts first by gathering context, checking known causes, and drafting incident summaries with recommended actions
• Own the path from shadow mode to paper trading to small live allocations, with explicit promotion criteria, kill-switches and rollback drills at every stage, and design the experimentation framework for A/B and interleaved parameter tests across venues and instruments
• Build defences against automated overfitting: a budget of trials per hypothesis, multiple-testing corrections (deflated Sharpe, false-discovery control), untouched holdouts agents cannot read, and independent re-execution of every claimed result
• Build eval suites with known-answer research tasks to track how often agents find signals, miss them or fake them, and route cheap models to extraction and classification while reserving frontier models for reasoning
• Stand up secure, on-prem GPU inference for open-weight models so proprietary data and strategy code never leave your infrastructure, and define which tasks may use external APIs and which must stay on-prem
• Ship all code through review and CI with no exceptions for agent-authored code or prompts, document architecture decisions and kill-switch conditions for operational continuity, and own systems end to end from design through production monitoring

What you bring

• 10+ years building production software or ML systems, including several years as a technical lead who set architecture for other engineers
• 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 building backtests, factor or signal models, or risk models, with deep familiarity of look-ahead bias, data snooping, regime change and transaction costs
• MLOps and CI/CD at platform level: you have built model registries, experiment tracking, automated validation and deployment pipelines, and treat reproducibility as a hard requirement
• Production streaming and analytical data systems experience with Kafka (or Redpanda) and ClickHouse, including schema design, partitioning, consumer lag and query performance at high volume
• Strong Python and SQL
• Comfort on Linux: debugging processes, memory, networking, I/O with perf and strace-level tooling, and running workloads on Kubernetes with infrastructure-as-code
• Experience shipping automated decision systems in regulated or capital-at-risk environments: kill-switches, audit trails, compliance review
• 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 such as 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 (perps, funding, on-chain venues)
• Market microstructure knowledge: order books, queue position, markouts, adverse selection
• MCP server design or open-weight model serving (vLLM, SGLang or similar) on your own GPUs
• Experience with Delta Lake, Dagster, MLflow, GitOps (Flux / Argo CD) or Terraform

What we offer

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

About Flowdesk

Flowdesk is a full-service digital asset trading and technology firm founded in 2020, providing liquidity solutions, OTC trading, and market infrastructure that connect digital assets to global markets with institutional-grade precision and scale. The company combines expertise from traditional finance, algorithmic trading, and blockchain technology to serve institutions, exchanges, and token issuers worldwide.