Your fractional quant

Stop spending senior research time on hypothesis testing.

Send me the research backlog. I’ll investigate it for you — rapid screening for every idea, deep investigation where the evidence warrants it, and a clear research conclusion at the end.

Send a research request See an example investigation

For portfolio managers, investment teams, small funds and family offices. Initial research results typically within 24 hours.

Fig. 1 — One screened idea, two honest numbersExample
OUT OF SAMPLE 2014 2024 3.0 1.0
As submitted (gross, same-bar fills) After costs, lag-1 execution

01 · The problem

More worthwhile research questions than researcher-hours.

Your team’s scarce resource isn’t AI tokens. It’s senior attention.

AI made research generation cheap. Validation, methodology, judgment and prioritization still consume skilled time. Somebody senior still has to decide what in the backlog is worth testing, test it properly, and decide what deserves more time or capital.

You don’t need another AI tool. You don’t necessarily need another full-time quant either. Give me the research backlog your team doesn’t have time to investigate. I screen ideas quickly, investigate the promising ones deeply, and give you a clear research conclusion.

02 · How it works

Rapid screening for every idea. Deep investigation where the evidence warrants it.

Not every idea needs the same depth. Everything gets screened; only what survives gets the full treatment.

Stage 1

Every idea

Initial results typically within 24 hours
  1. ScreenFormalize the question. Is it testable, and is it worth testing?
  2. TestImplement the baseline on clean data with realistic execution.
  3. ChallengeAttack the assumptions: lookahead, selection, costs, regime.
  4. VerdictReject · Investigate further · Incubate · Pursue.
Stage 2

Promising ideas

Where the screen says it is worth it
  1. Deep researchVariants, parameter neighbourhood, mechanism.
  2. RobustnessSensitivity, sub-periods, universe changes.
  3. OOSHeld-out windows and walk-forward where appropriate.
  4. CostsFees, slippage, capacity, financing.
  5. IncubationShadow-run and monitor before capital is discussed.

03 · What you can send

Anything sitting in the research backlog.

Click one to drop a realistic version into the request form. Rough is fine — I do the formalizing.

04 · What you get back

A research conclusion, not a collection of backtest metrics.

Every investigation ends in one of four decisions, with the evidence and reasoning behind it. The memo is short; the decision is explicit.

Reject
The idea fails the screen or the challenge. You stop spending time on it.
Investigate further
Something is there, but it needs a specific next test before anyone should care.
Incubate
Robust enough to shadow-run and monitor. Not yet a capital conversation.
Pursue
Survived robustness, OOS and costs. Worth a real allocation discussion.
§1

Research specification

Exact hypothesis, assumptions, universe, data and methodology.

p.1
§2

Evidence

Baseline results and the evidence that matters.

p.2
§3

Validation

Robustness, OOS, sensitivity, costs and failure analysis — where the evidence warrants it.

p.3
§4

Verdict

Reject, investigate further, incubate, or pursue — with reasoning and the recommended next step.

p.4

05 · Example investigation

What a finished memo looks like.

An illustrative investigation in the format every request comes back in. Names, numbers and dates are representative, not client data.

Send something to investigate
Research note · IllustrativeStage 1 → Stage 2

Range-position bounce on liquid index futures, after costs

Submitted as: an AI-generated strategy · Universe: US equity index futures · Horizon: daily

Specification. Close in the bottom quintile of the day’s range predicts next-session mean reversion, conditional on a longer-horizon trend filter. Lag-1 execution; realistic exchange and clearing costs.

Evidence. The submitted backtest’s in-sample Sharpe is not the result. After costs and lag-1 fills it compresses sharply. The held-out 2018–2020 window is positive but concentrated; the 2020 crash window is a wash.

Challenge. The parameter neighbourhood is unstable below the stated threshold. The trend overlay is doing real work — remove it and the effect disappears.

Verdict — IncubateShadow-run with the trend overlay and monitor live fills for a quarter. Do not allocate on the original specification.

06 · Beyond the backtest

Shadow incubation and ongoing monitoring.

Promising ideas do not stop at a backtest. They can move into shadow incubation: signals generated live on real data, positions tracked without capital, results compared against the research expectation. You see whether the edge shows up in the wild before it costs anything.

Incubated ideas are monitored on an ongoing basis — drift, regime changes, and breakdowns get flagged rather than discovered in the P&L.

07 · Why me

A quant researcher on retainer, responsible for the conclusion.

I am Vasilis Kotopoulos, a quantitative researcher with a systematic digital-asset background. I use AI heavily myself. That is the point: you are not paying me to write code an LLM can generate. You are paying for a quant researcher who uses AI as leverage while remaining responsible for the research conclusion.

Quant leadershipLed a six-person quant and data team.
Alpha researchSignals, regime models and portfolio construction for systematic strategies.
Robustness testingBuilt the tooling that stress-tests strategies before they see capital.
Infrastructure & deploymentBuilt and ran production trading infrastructure and shadow-incubation workflows.

08 · The plan

One plan. An ongoing research queue.

Fractional Quant

€500 / month

Continuous access to a fractional research function.

Send a research request
  • Ongoing research queue — send the backlog as it accumulates
  • High-throughput hypothesis testing: every idea screened
  • Deeper investigation of promising results
  • Robustness, OOS and cost validation where warranted
  • Clear verdicts: reject, investigate further, incubate, pursue
  • Initial research results typically within 24 hours

Need higher throughput, proprietary data integration, monitoring or deployment? Custom team retainers available.

09 · Send a research request

Have something sitting in your research backlog?

Send it over. I’ll show you how I’d investigate it. Or email hello@vkotopoulos.gr.

Send a research request