Please note: this role requires hands-on experience building and operating distributed batch compute for research or numerical workloads and production numerical Python. Applications without both will not be taken forward.

The firm

Our client is a systematic investment firm that already trades live across liquid markets on infrastructure built in-house on Google Cloud, automated end to end from data through execution. It is now building its dedicated research and engineering team in Dubai, and this is the first seat: the engineer who owns the research platform. The role is employed through the firm's DIFC entity, works directly with the founder, and carries full ownership of architecture and implementation.

The role

A working backtesting and optimisation stack exists today. Your mandate is to turn it into a distributed research platform that runs thousands of parallel backtests, simulations and parameter sweeps across assets, models and timeframes on elastic, Spot-capable cloud compute, with correctness, statistical validity and reproducibility built into the system rather than left to the researcher. The first sixty days are concrete: a 1,000-plus backtest sweep on pre-emptible infrastructure, with correct and complete results.

You will build the distributed research engine (scheduling, retries, failure recovery, result collection, monitoring and cost control on GCP Batch and Compute Engine Spot); the research data and feature layer (deterministic, versioned datasets on GCS and BigQuery in Parquet/Arrow, with point-in-time correctness, timestamps, historical universes, corporate actions and lineage); the optimisation and validation layer (large-scale parameter and model search with walk-forward, purged splits, strict holdouts, leakage prevention and multiple-testing-aware evaluation); and the reproducibility layer, where every material result is traceable to code, container, data, parameters and validation config, driven by a declarative strategy specification shared by research and production. At six months the core platform is in production use; at twelve it is the firm's standard research environment.

Essential

Around 4–7 years' experience, including personally building and operating distributed or batch compute for research, simulation or large-scale numerical workloads on GCP or AWS, at real scale, with Spot or pre-emptible capacity. Strong production Python with substantial numerical and data engineering (NumPy, pandas or Polars, Arrow/Parquet). Cloud data engineering on a columnar warehouse and object-store data lake (BigQuery/GCS or the AWS equivalents). The ability to become productive in a substantial statically typed codebase: the backtest engine is C# (QuantConnect Lean); professional C# is ideal, and concrete experience extending or debugging a substantial Java or C++ codebase is accepted in its place. A working understanding of research correctness in systematic trading: look-ahead and survivorship bias, walk-forward validation, overfitting and multiple testing, transaction costs. Willing to relocate to and work from Dubai.

Preferred

Large-scale hyperparameter optimisation (Optuna, Ray Tune or similar). Experience designing declarative specifications that drive pipelines or experiments. Background in a systematic fund, proprietary or HFT trading firm, or established crypto trading firm; large-scale ML-infrastructure or scientific-computing teams are also relevant where the compute experience is evidenced.

This is not

A DevOps, SRE, generic data-engineering or cloud-platform role, a low-latency execution role, or a quant researcher seat, and it does not suit engineers whose research experience is writing strategies on top of someone else's platform or single-machine C#/C++ systems without cloud-scale compute.

Package

AED 456,000–480,000 base (approx. US$124,000–131,000), tax-free in the UAE. Discretionary bonus with a 20–40% target against quarterly and annual deliverables. Equity-linked incentive (profit units) from six months. DIFC employment visa, relocation support, and a notice-period buy-out for the right candidate.

Process

Three stages with the founder: a 90-minute technical deep-dive including a short C# code-reading exercise; a live working session of around three hours on video (a market-data pipeline component and a result-collection design for distributed backtests on pre-emptible capacity; no take-home, no algorithm puzzles); and a founder conversation on mission and package. Offers are made within three business days of the final stage.

To apply

Send your CV to tabby@platinumandpartners.com quoting REF 1065. All conversations are confidential.

Similar jobs