Fraud & Risk Analytics
Detecting behavioural patterns, suspicious activity, abuse signals, and operational risk.
Independent Analytics Consultant · Analytics Leader · Payments · Fraud · BI
Sergey Kaminov
I help organisations solve difficult analytics, payments and fraud problems — turning fragmented commercial, behavioural and operational data into trusted decisions, controlled risk and measurable outcomes. Independent consulting and fractional analytics leadership, with 15+ years and hands-on teams of 15–25 when a problem needs them.
From analysis to organisational action.
Strategy · Leadership · Data · Risk · BI · AI
Experience
15+ years
analytics, BI, commercial systems and transformation
Leadership
15–25-person project teams
analysts, developers, project managers and domain specialists
Measured impact
~20% uplift · ~10% saving
multi-market marketing and payment-processing outcomes
Research edge
AI coherence
independent work on long-horizon stability and repair
00
Choose the path that matters to you
Three ways to engage: commercial project work and fractional leadership, permanent leadership roles, and independent research.
For organisations
Consulting, fixed-scope projects and fractional analytics leadership.
Diagnostics, investigation sprints, BI/KPI resets and a fractional lead — most start with one tightly scoped, low-risk engagement.
For hiring managers Permanent or long-term analytics leadership roles.Head of Analytics, Head of BI / Insights, senior management and data-led transformation.
For research collaboration AI coherence, self-repair and long-horizon stability.The CRepair work, replication notes and the open research archive.
01
Analytics & Decision Systems
I help organisations turn complex behavioural, financial, and operational data into decision systems.
Across more than 15 years, I have worked across data analytics, business intelligence, fraud prevention, payment behaviour, customer systems, reporting architecture, dashboards, and AI-assisted analytical workflows.
My work focuses on the behaviour of systems: how users act, how risk emerges, how incentives shape outcomes, and how data can be structured to improve commercial, operational, and strategic decisions.
Professional focus
Analytics leadership · BI & insights · Digital transformation · Fraud & payments · Commercial decision systems · AI-enabled workflows
Payments & Behaviour
Understanding payment flows, user behaviour, transaction patterns, and commercial outcomes.
BI & Reporting Architecture
Building dashboards, analytical frameworks, reporting systems, and decision-support structures.
AI-Assisted Analytics
Using modern AI tools to accelerate analysis, reporting, quality checks, documentation, and decision workflows.
Worked example · illustrative
See the example ↗
From messy data to a decision
A visual walk-through of how raw, broken data becomes a quantified business action — messy input → coherent structure → the signal → a rule → measured impact. Switch between fraud & payments and retention & churn.
Analysis · AI & the US job market
Open the seismograph ↗
AI Labour Seismograph
An interactive model of AI task-exposure, BLS labour-market signals and the full cost of automation — supervision, integration, retries and error risk, not just tokens. It tests where AI actually beats fully-loaded labour, and where the "AI replaces everyone" story breaks down on the economics.
Available for independent consulting, fixed-scope analytics projects and fractional analytics leadership — plus selected permanent leadership roles. See how to work with me →
02
Leadership & Business Transformation
The value is not only in producing analysis. It is in setting direction, organising delivery, creating trust in the numbers and moving the organisation from insight to action.
Strategy & Operating Model
Define the decisions, metrics, ownership and delivery roadmap that an analytics function needs to create business value.
People & Delivery Leadership
Lead multidisciplinary project teams of 15–25, aligning analysts, developers, project managers, specialists, suppliers and client-side owners around one delivery plan and quality bar.
Executive Influence
Translate complex evidence into commercial choices, trade-offs and leadership-ready action rather than static reporting.
Transformation & Adoption
Unify fragmented definitions, systems and teams around a trusted measurement framework and an operating rhythm that lasts.
Project teams of 15–25
Consistently led multidisciplinary UFS Consulting teams combining outsourced developers, project managers, analysts, technical specialists and client-side owners.
Owned delivery from diagnosis and roadmap through workstream coordination, adoption and verified outcome.
Seven-country KPI rollout
Designed and led adoption of a common campaign-performance framework across seven TUI markets, aligning definitions, data and executive reporting.
Approximately 20% group-wide campaign-performance uplift.
Consulting under commercial scrutiny
Led analytics transformation, KPI-system rollout, card-acquiring and payments optimisation, and fraud-prevention implementation where value had to withstand client and financial scrutiny.
Delivery connected analysis, technology, operational change and a commercially defensible result.
Commercial team of 15
Managed a 15-person regional sales team at Nestlé using structured KPIs, coaching, negotiation and clear performance accountability.
Consistent target delivery in a demanding commercial environment.
Best-fit leadership roles
Hiring route
Head of Analytics · Head of BI / Insights · Analytics Manager / Senior Manager · Data & Analytics Transformation Lead · Commercial, Fraud or Payments Analytics Lead.
See the focused leadership profile, evidence and first-90-days approach. ↗
03
Selected Analytics Work
Sanitised commercial cases plus public reference builds — showing how I move from governed data to evidence, controlled action and verified outcome.
Sentinel
An end-to-end payment-risk case study translating fraud scores into operational approve, review and decline decisions.
Chronological validation, fraud capture, customer friction, revenue exposure, manual-review workload and decision-policy design using synthetic data.
MarginGuard AI
A governed revenue-investigation system that validates data, uses typed analytical tools, records evidence, gates consequential actions and verifies outcomes.
Interactive zero-cost replay on the site; executable PostgreSQL, dbt, FastAPI and audit implementation in the repository.
Run verified replay →Next Best Offer
A calibrated propensity and expected-margin policy selecting three eligible travel ancillaries for each campaign.
+18.34% simulated margin vs Popular Top-3 and +8.74% vs a frozen rule policy; reported honestly as Outcome C because Popular remained stronger on NDCG@3.
Explore public evidence →TUI · Group Marketing
Built group-wide CRM campaign reporting and analysis, and rolled out unified KPIs across 7 countries — so every market read campaign performance the same way.
Group-wide uplift in campaign performance (≈20%).
Read sanitised case →Betfair · Payments Analysis
Built the payments view management used to steer card processing — forecasting and budgeting models plus automated KPI dashboards that put decisions on live numbers.
~10% cut in interchange fees. Separately, traced a single misclassified referral-fee stream on corporate accounts — the correction lifted the company's reported profit margin from 10% to 11%.
Read sanitised case →UFS Consulting · Fraud & Payments
Led multidisciplinary project teams of 15–25 across NDA-protected analytics transformation, KPI rollout, card-acquiring and payments optimisation, and fraud-prevention implementation programmes.
Coordinated outsourced developers, project managers, analysts, specialists and client stakeholders from diagnosis through adoption; outcome-based work required value to be quantified, verified and commercially defensible.
Read sanitised case →
How I create value
Professional profile
Clarify the decision → structure the data → build the metric layer → automate the reporting loop → expose the risk, behaviour or commercial signal leadership can act on.
Download CV for full experience, tools, certifications and employment history. ↗
04
Independent Research
How do intelligent systems maintain coherence under uncertainty?
Alongside my analytics career, I conduct independent research into AI coherence, alignment, complex adaptive systems, consciousness models, and the stability of intelligent systems under uncertainty.
It's the same question that drives my day work — how complex systems hold together, and how to make that visible — followed past the dashboard into how intelligence itself stays coherent.
Research credibility
ORCID record · DOI-linked preprints · public code and datasets · benchmark design · cross-model replication · explicit claim boundaries.
Positioning
Independent researcher, not an institutional title claim. Credibility is built through inspectable methods, reproducible artefacts and collaboration.
Interactive explainer · companion to the consolidated preprint
Can an AI repair its own incoherence?
Play with the detect → repair → verify → stability loop and watch why verification — not detection — is the binding constraint. Updated around the consolidated CRepair preprint on verification-first self-repair for LLM agents.
Explore the CRepair explainer →
uncertainty
coherence
repair
stability
Start here
Collaboration & external work
- CRepair consolidated preprint — published. Verification-first self-repair for LLM agents, consolidating the benchmark, runtime wrapper, cross-model replication and component ablation into one engineering-facing preprint.
- Co-authored paper — in final review. Comparative characterization of embedding-coherence metrics under progressive semantic perturbation, with Javier Marín (Universidad Europea, Madrid) — in a final pass with the co-author ahead of submission.
- Judge-validation study — in progress. Human spot-checks against the LLM-as-judge coherence score, to ground the CRepair metric beyond automated scoring.
- Open code & data. CRepair benchmark and intervention protocols published openly so results can be independently replicated.
AI coherenceAlignmentComplex adaptive systems
CRepairConsciousness modelsDecision integrity
Long-horizon system stability
Selected publications
- Verification-First Self-Repair for LLM Agents: A Detect–Repair–Verify Benchmark, Runtime Wrapper, Cross-Model Replication, and Component Ablation 2026 · Consolidated preprint, Zenodo · 10.5281/zenodo.21452222
- CRepair Wrapper Improves Structural Self-Repair Across Three LLM Families: A Cross-Model Replication Study 2026 · Preprint, Zenodo · 10.5281/zenodo.20432173
- Detecting and Repairing Coherence Failures in Long-Horizon AI Agents 2026 · Preprint, Zenodo · 10.5281/zenodo.20283434
- Foundations of Coherent Artificial Cognition: A Structural-Dynamical Framework, Global State Architecture, and the A-Test Protocol 2026 · Preprint, Zenodo · 10.5281/zenodo.18445064
Also in the programme: CRepair benchmark foundation · Structured Runtime Intervention (CRepair pilot) · Cross-model replication · Component ablation · Coherence-per-Watt · CRepair Benchmark v0.2 (open software). Full archive — 20+ works on ORCID ↗ · deeper research at fractality.co.uk ↗
05
Labs — Visual Systems & Data Instruments
Experimental interfaces for exploring knowledge networks, planetary data, live signals and complex systems in motion.
Knowledge Network
WikiSphere
Wikipedia APIGraph UIExport
A knowledge-map interface for exploring Wikipedia as a living graph.
WikiSphere turns encyclopaedic knowledge into an explorable network, revealing connections between ideas, topics, and concepts.
Open WikiSphere ↗
Planetary Signals
TREMOR
Live feedsUSGS/NASAGlobe
A live planetary signal instrument visualising earthquakes, storms, fires, volcanoes, and floods on a rotating Earth.
TREMOR uses open scientific data feeds to show the Earth as a living system of seismic and environmental signals.
Open TREMOR ↗
Comparative Mythology
Tree of Religions
GenealogyLayered viewCanvas
A genealogical map of the world's faiths — documented lineage on one layer, and hidden, contested, mythological connections on another.
Tree of Religions traces who descended from whom across millennia, then overlays the contested and mythological links traditions rarely admit.
Open Tree of Religions ↗
06
Latest Signals
Recent public work: practical analytics, research, and visual instruments in one living body of work.
Analytics
Next Best Offer · Travel Ancillaries
Calibrated propensity, expected-margin Top-3 ranking and an honest frozen test result with public Kaggle and GitHub evidence.
Analytics Sentinel · Payment Risk AnalyticsFrom fraud scores to approve, review and decline decisions — a public Kaggle notebook with outputs and policy analysis.
Research CRepair consolidated preprintVerification-first self-repair for LLM agents: benchmark, wrapper, replication and ablation.
Analysis AI Labour SeismographUS labour-market signal and AI replacement-economics model.
Lab TREMORLive planetary signal instrument for earthquakes and hazards.
Book Human 2.5A phenomenological book on ground, perception and hidden supports.
07
Books & Essays
Books and essays on coherence, artificial intelligence, consciousness, civilisation, memory, and the preservation of meaning.
Independent work exploring coherence, consciousness, artificial intelligence, and complex systems.
A continuation of the Fractal Threads project, expanding the inquiry into coherence, consciousness, AI, and meaning.
A civilisational and systemic volume on coherence, institutions, AGI, society, ethics, and the noosphere.
A literary project about memory, forgetting, transmission, and the preservation of meaning after history is gone.
A careful phenomenological project about the hidden supports of ordinary human experience: ground, body, time, perception, memory, and self-location.
A children's comic about kindness and memory — the story of The Little Thread retold in pictures.
Selected essays
- When Intelligence Becomes Infrastructure Essay · Substack · Jul 2026
- We Do Not Live in the Present Essay · Substack · Jun 2026
- Why Intelligence Fractures Essay · Substack · Dec 2025
- Human 2.5: We See in 3D But We Live by the Ground Essay · Medium
- We Have Made This Mistake Before Essay · Medium
- The AI Genie, the Lucifer Paradox, and the Geometry of Awakening Essay · Medium
08
Field Notes
Beyond analytics and research, I am drawn to activities that make invisible systems tangible: sailing, skydiving, mountains, watches, electronics, and sound.
They all involve the same quiet discipline: reading signals, understanding constraints, and finding coherence under uncertainty.
Sailing
Skydiving
Mountains
Watches
Electronics
Sound
08
What colleagues and leaders say
“Sergejs was in my data/support team at Betfair. He was a joy to work with, providing insightful data with a great narrative. Sergejs is always happy to try new things and work outside of his comfort zone.”
“Sergejs and I worked together at Betfair several years ago. Sergejs was always very professional, and it was a pleasure to work with him!”
Professional recommendations from former colleagues and managers — not client testimonials. Verify on LinkedIn →
09
Contact
For consulting and project work, fractional analytics leadership, permanent leadership roles, research collaboration, or a considered conversation.
Primarily available for independent consulting, fixed-scope analytics projects and fractional analytics leadership across payments, fraud, BI and commercial decision systems — and open to selected permanent leadership roles (Head of Analytics, Head of BI / Insights, Analytics Manager or Senior Manager). See how to work with me →