About TradeVerge AI
Who you'd be joining.
TradeVerge AI builds verified, rules-based automation for the gold market. Our flagship product, Gold Core, is a MetaTrader 5 Expert Advisor with a public, third-party-verified track record on Myfxbook and thousands of members inside our live desk. We are a small, remote-first team that prefers evidence to opinion: every claim we publish is auditable, every release is versioned, and every support answer is written by someone who understands the code.
Research sits at the front of everything we ship. The desk owns the data pipeline, the validation protocol and the parameter sets that reach live accounts. Nothing is released on a hunch: each change carries a memo, an out-of-sample result and an expectancy range that support and marketing are held to.
Why this role matters
We are deliberately sceptical of black boxes in live trading. This role finds the narrow places where learning models beat rules, proves it, and keeps everything explainable to a member who asks why.
What you'll actually do
The day-to-day, not the job-spec version.
- Build regime-classification and anomaly-detection models over market and execution data
- Design evaluation that respects time ordering and avoids leakage
- Ship models behind clear guardrails with human-readable explanations
- Monitor drift and retire models that stop earning their place
The team and how you'll work
You'll work inside the research desk, reporting to the lead for research, and partner day to day with MQL5 Engineering, Product, Support, Compliance.
We are async-first across European and American time zones. Expect a short daily written check-in, one team sync a week, and long stretches of uninterrupted focus in between.
Decisions are written before they are made. If a change reaches members, there is a document explaining why, what we expect, and how we'll know if we were wrong.
Tools you'll live in: Python (pandas, numpy, polars) · Jupyter / Quarto · Postgres + TimescaleDB · MT5 Strategy Tester · Git.
Your first 90 days
What a good start looks like — we'll agree this with you in week one.
- Complete onboarding with the research desk: environments, access, and a walkthrough of how Gold Core actually behaves on a live account.
- Read the last two release memos end to end so you understand the standard of evidence we hold ourselves to.
- Ship one small, real improvement in your area — scoped with your manager in week one.
- Meet everyone you'll work with regularly across MQL5 Engineering, Product, Support.
- Take full ownership of a defined area and become the person the rest of the team asks about it.
- Deliver your first substantial piece of work end to end, including the written rationale behind it.
- Identify the two weakest things in your area and put a fix for one of them on the roadmap.
- Run your area independently, with your manager reviewing outcomes rather than steps.
- Improve one process or piece of tooling so the next person does the job faster than you did.
- Agree the next two quarters of objectives with clear, measurable success criteria.
How success is measured
No vague reviews — these are the things we'll actually talk about.
- Live expectancy tracks modelled expectancy within the published band
- Every shipped parameter change is backed by a reproducible research memo
- Regime drift is flagged before members notice it, not after
- Rejected ideas are documented as clearly as accepted ones
Who we're looking for
The full qualification list for this role. Very few people tick every line — apply if the essentials fit.
The core of the role — we read these closely.
- 5+ years of professional experience in research or a directly adjacent discipline
- Demonstrable ownership of work at the scope described above — ideally on a product used by paying customers
- Production ML experience with time-series or sequential data
- Rigorous evaluation discipline and healthy scepticism of your own results
- Ability to explain a model to a trader without jargon
- Strong Python for research work: pandas/numpy/polars, vectorised backtesting, and reproducible notebooks
- Applied statistics — hypothesis testing, bootstrapping, Monte Carlo, multiple-comparison and overfitting controls
- Time-series handling at tick and bar resolution, including gaps, rollovers, session boundaries and survivorship issues
- Comfort with SQL and columnar/time-series stores for multi-year market datasets
- Version-controlled research (Git) with results that another person can re-run from scratch
- Working knowledge of FX/metals microstructure: spread, slippage, swap, liquidity by session
- Experience designing walk-forward and out-of-sample protocols for systematic strategies
- Ability to size and stress-test risk: drawdown distributions, exposure limits, tail scenarios
We hire on evidence; credentials are a signal, not a gate.
- Degree in a quantitative discipline (maths, physics, statistics, CS, financial engineering) or demonstrable equivalent work
- CFA, CQF, FRM or comparable certification welcomed but never required
- Writes clearly and thinks in public — decisions are documented before they are made
- Comfortable being wrong quickly: you look for the evidence that would disprove your own view
- Self-directed across time zones, with reliable follow-through and no need for chasing
- Holds a high bar on member-facing quality, including when nobody would notice the shortcut
- Treats compliance, honesty in marketing claims and risk disclosure as part of the craft, not overhead
- Legally able to work in United Kingdom (London (hybrid, 2 days in office)) without sponsorship at this time
- Available for full-time work and a reliable home or co-working setup
- Fluent professional English, written and spoken
- Willing to complete a paid, time-boxed craft exercise as part of the process
Nice to have
Genuinely optional — strong candidates rarely tick every box.
- Finance or trading experience
- Bayesian methods
- MLOps tooling
Compensation and benefits
£90k – £125k — the band is set by level, published up front and paid the same regardless of how hard you negotiate. Reviewed annually, plus an out-of-cycle review if your scope changes.
- Fully remote, async-first — we optimise for deep work, not meeting attendance
- Competitive base plus performance bonus tied to shipped outcomes
- Private medical cover (US/UK) and a health stipend elsewhere
- £/$2,000 annual learning budget for research, data and conferences
- Top-spec hardware, multi-monitor setup and VPS/data subscriptions covered
- 28 days paid leave plus local public holidays
How hiring works
Five stages, roughly ten days end to end.
- 01Intro call (30 min) — your background, what you want to build next
- 02Craft interview (60 min) — a real problem from our backlog, no whiteboard trivia
- 03Paid take-home or working session (4–6 hrs, compensated)
- 04Team conversation with two people you'd work with daily
- 05Offer, usually within 10 days of the first call
We are looking for honest statistics and the willingness to kill your own idea. Everything in our process is designed to surface that rather than test recall.
The craft interview uses a real, recent problem from our backlog. You may use any resource you would use on the job, including documentation and AI tools — we care how you think, not what you memorised.
Take-home work is paid at a fair hourly rate and capped so it never eats a weekend.
You will get a decision, with reasons, after every stage. We do not ghost.