Aquatic Capital Management: Careers, Jobs & Salary Guide

Aquatic manages systematic investment strategies and describes research, engineering and risk management as central to its business. Its official website identifies founder Jonathan Graham and his previous leadership of Citadel’s Global Quantitative Strategies business. The legal disclosures identify Aquatic Capital Management LLC as the US investment adviser and Aquatic Capital UK LLP as a UK affiliate.

Official navigation: company website · open jobs and applications · regulatory profile. The company site lists Chicago at 225 W Randolph Street, New York at 461 Fifth Avenue and London at 1 Angel Court.

The company does not publicly disclose a complete strategy catalogue, holding periods or proprietary model architecture. A job mentioning machine learning does not establish that every strategy uses deep learning. A software role mentioning latency does not make every investment strategy high frequency.

Check the individual vacancy for its advertised location. A company office in a city does not mean every role is offered there.

Public directories sometimes disagree because they measure different entities or reuse older data. The following figures come from identifiable records rather than a blended estimate.

These are dated disclosures. The filing date is not a live employee count or a daily asset valuation. Four reported accounts also do not mean four individual investors. Use the IAPD profile to find a newer filing when available.

Regulatory AUM, fund gross asset value and reportable 13F holdings are different measures. Do not substitute them, add overlapping master/feeder fund values or interpret AUM growth as investment returns.

The official careers board showed these seven openings on 3 October 2026. Each role name links to its employer-authored description. Recheck availability before submitting an application.

Competitions and previous trading experience are advantages in several research adverts, not universal requirements or substitutes for the required skills.

Some application forms offer more location choices than the vacancy’s title lists. Use the advertised role location as your starting point and ask recruiting about any discrepancy. In particular, the research internship explicitly limits its locations to Chicago and London.

Specific full-time postings disclose base salary. These are advertised ranges, not observed median salaries, accepted offers or total compensation.

The full-time adverts describe possible additional compensation, including a discretionary bonus. They do not disclose a typical bonus amount. The experienced-research advert’s dollar range should not be relabelled as a confirmed London salary, converted into an implied GBP offer, or generalized to every employee.

When discussing an offer, separate base, guaranteed versus discretionary bonus, payment timing and relocation terms. Ask for the office-specific package and currency. An application question about your total compensation expectations does not disclose the employer’s expected total pay.

Choose the problem you want to own. Research owns evidence for a forecast or strategy; production engineering owns its path into reliable live systems; infrastructure engineering owns the computing environment. Prepare for the advertised responsibility.

No separate Machine Learning Engineer opening was listed in the reviewed board. The experienced research advert includes statistical and machine learning models; that is evidence of research responsibilities rather than a new job title.

The production role emphasizes releases, rollback, orchestration and monitoring. The systems role explicitly mentions Slurm/Ray, hybrid on-prem/cloud computing, GPU workloads and storage. Those details can guide a relevant conversation; they do not reveal the firm’s complete technology stack.

This is an independent preparation exercise based on the skills the research adverts request. Choose a manageable dataset and question. The goal is to make every important decision inspectable.

1. Specify the question and timing. Write what you predict, the horizon and exactly when each input becomes available. For a daily-return study, distinguish a value observed after the close from a signal you could have acted on before that close. Document missing values, revisions and the universe selection.
2. Create a baseline before a complex model. Use a simple forecast or regularized linear model suited to the question. Keep the dataset, target and evaluation periods identical when comparing a more complex model. An improvement should survive a fair comparison, not a change in the experiment’s conditions.
3. Freeze a chronological evaluation design. Keep a final period untouched while choosing transformations and hyperparameters on earlier data. Fit preprocessing on the training period only. If labels use future observations, prevent those observations from crossing the boundary into the training set. Log every variant you try.
4. Report prediction quality and economic assumptions separately. Show the metric relevant to your target and how it varies across periods. If you add a simulated trading layer, state when positions change, how exposure is constrained and which costs you assume. A forecast score is not a tradable net return. A synthetic exercise is not evidence of live profitability.
5. Investigate a failure deliberately. Compare results by period, group or volatility environment. Try a delayed input, a missing-data case or a higher assumed cost. Explain whether the weakness comes from the signal, the data, the evaluation or the implementation. Preserve an unfavorable result instead of quietly removing it.
6. Package the study for reproduction. Include a README, environment requirements, data instructions, deterministic entry point, tests for critical transformations and a brief discussing the conclusion and limitations. Respect data licences and keep private employer work out of a public repository.

For ML, add an ablation table: baseline; additional features; more complex model; each tested under the same split. Explain whether gains justify extra training cost and maintenance. A carefully explained thesis or replication can supply comparable evidence for early-career or PhD applicants.

A student exercise and a senior application need different evidence.

For software engineering interns, a small data-processing service can show algorithms, interfaces and testing. Define behavior for duplicated, delayed and malformed records. Measure throughput and memory on a fixed workload. Test correctness before optimizing, then document one measured improvement and its trade-off.

For production platform applicants, demonstrate an immutable build artifact and a controlled release path. Add configuration validation, a health check, logs and metrics, and a rollback procedure. Inject a failure such as a broken dependency or an interrupted deployment; record whether detection and recovery work. Explain how another engineer can operate the system after you leave.

For systems applicants, use real experience where you have it: cluster scheduling, storage bottlenecks, Linux debugging or capacity planning. Explain the workload, symptoms, measurements, competing explanations and outcome. A small local demonstration may support the discussion, but it cannot stand in for the systems advert’s ten-year experience requirement.

Across these directions, report measured changes with hardware, input size and test conditions. A successful benchmark alone does not establish production reliability.

Confirm the posting’s degree/experience category and location, and save its requirements with the date. A useful CV project bullet states the question or system, your contribution, the method and a result with its limitation.

The reviewed forms ask for a GitHub username and a public SSH key added to that account. Several explicitly say this is required to advance in the process. Read the instructions on your chosen application; a public key is different from a private key. They also ask about employment authorization and sponsorship needs. A question about sponsorship is not a guarantee that sponsorship will be offered.

Ask recruiting about the interview format for your exact role, permitted coding languages, project discussion expectations, office arrangements and timeline. For research, useful later questions concern how ideas are reviewed and how researchers collaborate with engineering. For engineering, ask about ownership, incident responsibilities and how platform improvements are measured.

Aquatic’s official board lists recruiting@aquatic.com for experienced professionals and campusrecruiting@aquatic.com for students. The public material reviewed does not establish a universal number of rounds, required take-home exercise, question bank or response deadline. Prepare the relevant skills and verify the actual process with the employer.

Use this schedule to audit existing work and identify gaps; adapt the pace to your starting point.

If the audit exposes a foundational gap, slow down and study it. The QuantSteps research path helps organize statistical and programming foundations. Regression diagnostics provides further model-assumption practice with Pro access; these are independent learning resources, not an Aquatic curriculum or hiring assessment.

The free Returns & risk lab offers a smaller starting point: synthetic prices, return calculations, compounding, drawdowns and a written brief. Apply through Aquatic’s official channel when your eligibility and evidence match the advertised opportunity.