Accepting new engagements
We reproduce backtests, test forecasting and risk models, and deliver the findings memo, reproducible notebook, and remediation path your team can inspect.
Start with the decision, the model, and what must hold up under review. Scope and fee are agreed in writing before work begins.
StatGazer / public sample work product
Illustrative · synthetic data · not a track record
Sample deliverable
A deliberately flawed synthetic strategy is rebuilt on point-in-time inputs so the decision-maker can see what survives.
| Metric | Reported | Reproduced |
|---|---|---|
| Sharpe (annualized) | 1.8 | 0.7 |
| Annual return | 14% | 6% |
| Max drawdown | −9% | −21% |
| Hit rate | 58% | 52% |
Look-ahead in the signal pipelineA feature was joined on report date instead of its as-available date.
Founder-led review
Founder · Data scientist · Financial engineer · Your direct reviewer from scoping through handoff
01section Work product
The product is the evidence trail: a findings memo that states what changed, a reproducible notebook for the calculations it contains, and a remediation path your team can act on. Client work stays confidential; the memo sample and notebook below are separate synthetic examples of the deliverable standard, not client results.
# randomness is explicit
RNG_SEED = 20260609
df["momentum_20"] = (
df["asset_ret"]
.rolling(20).mean().shift(1)
)
# finding: future data enters here
df["leaking_future_mean_5"] = (
df["asset_ret"]
.shift(-1).rolling(5).mean()
)The full public notebook records its seed, data-generating process, leakage check, walk-forward split, and findings.
Evidence: the feature contract marks the input unavailable at trade time.
Action: remove it, rebuild on point-in-time inputs, and rerun the full validation path.
02section Engagements
Choose the smallest review or build that can answer the question. The model audit is the flagship path; adjacent work uses the same evidence standard.
Flagship · E.01
An independent technical review before you allocate, present a result to an investment committee, or inherit a model you did not build.
E.02
A model built from your data, validated out of sample, delivered with uncertainty stated and a path to monitoring.
E.03
Independent review of the model, pipeline, data lineage, monitoring, and governance before a vendor or system becomes operationally important.
03section Review path
Each checkpoint produces something your team can inspect, challenge, and keep. No unexplained score and no junior handoff between the question and the answer.
Define the model, data, decision, constraints, and what must be defensible to the reviewer who matters.
Written scope + evidence planRebuild the path from raw or minimally processed inputs where access allows, making timing, assumptions, exclusions, and manual steps explicit.
Reproduction notebookTest leakage, regimes, out-of-sample behavior, uncertainty, monitoring needs, and remaining limitations.
Findings memo + remediationWalk your team through the code, conclusions, limits, and the triggers that should prompt another review.
Code + report + handoffNDA-first, minimal extracts where possible, founder-only access unless agreed otherwise, and no third-party AI training on client data.
You own the agreed deliverables. The engagement agreement states ownership and license terms; the written scope records access requirements, deliverables, timing, and fee before work starts.
Technical model review and research consulting — not a financial-statement audit, regulatory assurance, investment, legal, or tax opinion.
04section Practical details
If your process needs a different NDA, onboarding packet, or review boundary, raise it during scoping.
Yes. We sign an NDA before reviewing confidential data and are glad to work under yours. Where possible we work inside your environment or on a minimal extract. Access is founder-only unless otherwise agreed, and we do not use client data for third-party AI training.
You own the deliverables produced for the engagement, including agreed code, models, notebooks, and reports. We retain pre-existing methods, tooling, and general know-how. The engagement agreement states the exact ownership and license terms.
Productized engagements may publish a fixed fee for a defined scope. Custom work is priced after scoping: a fixed fee where work is well-defined, or a written rate and estimate where open-ended. No work starts, and no fee is charged, outside an agreed written scope. Timing is set by the published product scope or the custom written scope.
An outside reviewer gives the internal champion a defensible second read from someone who did not build the original model. Your team keeps ownership and context; StatGazer adds a reproduction record, findings memo, and remediation trail that can travel to an IC, risk committee, or vendor review.
No. StatGazer provides technical model review, validation, research, and engineering consulting. It is not a CPA firm, broker-dealer, investment adviser, or regulatory assurance provider, and its deliverables are not financial-statement audit opinions or regulatory assurance reports.
05section Next step
Send a high-level description without confidential data. We will tell you the shortest honest path to an answer and whether StatGazer is the right reviewer.
Replies within 24 hours · Direct founder line hello@statgazer.com