Director of Data Science & Engineering, Fraud Platform
for each section heading. Use - for any list of duties/requirements/benefits. Use
for prose.
We must not repeat job title as heading. No Job ID.
We must preserve all substantive information; do not invent or summarize away detail. Must keep details.
We need to extract content precisely.
Let's parse:
About the role paragraph: Starting after "About the role". The text:
"This is a high-visibility, high-accountability role — outcomes are reported to C-level and the board, and the role owns both the data strategy and the model performance behind systems that directly protect revenue and customer trust. The mandate in one line: own what data is generated and how it is generated and consumed across the Identity and Financial Crime domains. Almost every consequential decision Zepz makes about a transaction — is this person who they say they are, is this payment safe, is this account compromised — is a decision made on data. Identity is our single largest producer of that data. Financial Crime is our single largest consumer of it. Both sit in the same engineering organization, and today no one owns the path between them. The result is a permanent tax on everything downstream: problems discovered late, by the people least able to fix them at source. This role exists to remove that tax. You will lead a team of 25+ Data Scientists, Machine Learning Engineers and Analytics Engineers across the Identity and Financial Crime domains"
That's the About the role prose.
Then "What you will own" bullet list:
- Own the definition, generation, quality and consumption of data across the Identity and Financial Crime domains — what gets produced, what it means, what guarantees come with it, and who is accountable when those guarantees are not met.
- Make data quality an engineering discipline rather than a downstream cleanup exercise: producer/consumer contracts, freshness and integrity monitoring, lineage, a named owner for every critical dataset, and a real path from a detected anomaly to a fix at source. You'll be measured on whether downstream teams stop discovering data problems by accident.
- Own the models and analytics behind how Zepz detects fraud and financial crime — feature engineering, model development, deployment, monitoring and retraining. You own model performance in production, not just model development.
- Lead a mixed organization of Data Scientists, Machine Learning Engineers and Analytics Engineers. These are three different crafts with three different career paths, and leading them well means understanding all three rather than favoring the one you came from.
- Partner as a peer with the Engineering Manager who owns the fraud detection platform and reports directly to the VP Engineering. Your models run on their platform: you own model outcomes, they own platform outcomes, and you jointly own the seam between them.
- Partner with our central data function, which owns core data infrastructure and platform tooling. You are a demanding customer and a collaborator, not a replacement — you build domain data products and quality practice on top of what they provide, and feed real requirements back to them.
- Set standards with Identity engineering teams who do not report to you. Much of your impact depends on changing how data is produced upstream, which is influence work: making the cost of poor data visible with evidence rather than assertion, and making it easy for those teams to do the right thing.
- Create high-performing teams with a bias toward shipping without sacrificing reliability or trust in the numbers. Use technical and data debt as a tool to move faster, while ensuring your teams have the space to build it right once they know what the right thing to build is.
- Hire, retain, and level up superlative talent to ensure our team remains world-class as we continue to scale and launch into new markets.
- Roll up your sleeves when models degrade or data breaks. You'll need to discuss technical solutions, review work in depth, and help execute strategy with your leads.
- In your first few months, you will:
- Establish an honest, evidenced baseline of data quality across the identity-to-financial-crime path — what breaks, how often, who finds out, and what it costs in downstream hours and model degradation. Bring the number, not the anecdote.
- Agree a written ownership model with the central data team and the Identity engineering leads: who owns which datasets, what the contracts are, and what happens when one is broken.
- Instrument the critical path, so data issues are detected by monitoring rather than by an analyst noticing that something looks odd.
- Get to know the models — how they perform, how they degrade, how retraining actually works today, and where risk is concentrated.
- Assess the team: identify the gaps, the flight risks and the people ready for more, and start hiring against a plan rather than against a vacancy.
- Promote a culture of ownership, accountability, and shipping.
Note: The bullet list includes a sublist under "In your first few months, you will:" which is also bullet points. We'll treat as part of responsibilities.
Now "What you bring to the table" bullet list:
- You have a track record of leading a data organization of comparable size (roughly 15–25 people) and of supporting the professional development of managers and senior-level ICs.
- Track record of leading mixed technical teams — Data Scientists, ML Engineers, Analytics Engineers — with genuine credibility across all three crafts.
- Demonstrable ownership of data quality at scale: contracts, observability, lineage, semantic consistency — with evidence of the downstream impact, not just the tooling you installed.
- Fluency with modern lakehouse and analytics engineering practice: Databricks (or equivalent), dbt-style transformation, feature store concepts and real-time feature serving, and the discipline of treating datasets as products with owners and SLAs.
- Experience with production ML in a real-time context: model monitoring, drift detection, and retraining pipelines.
- Proven experience with AWS at scale.
- Direct experience leading FinCrime/Fraud or Identity engineering or data science teams (not just adjacent risk/security).
- The ability to get outcomes from teams that don't report to you. We'd rather hear about a time you changed a producer team's behavior than a time you escalated.
- At least 8+ years of technical experience in a hybrid IC/management role.
- You can attract and retain top talent and lead teams with diverse skill sets.
- You work well with Engineering and Product leaders and are comfortable building effective relationships with folks outside of the Prod-Eng spheres.
- You have exceptional judgment; we're scaling quickly — you'll be required to make good tradeoffs.
Then "FinCrime & Identity Domain" paragraph:
"Bring prior experience leading FinCrime/Fraud or Identity teams — understanding of typologies (card fraud, ATO, first-party fraud, synthetic identity, money mule networks), the regulatory context, and how detection strategy trades off against customer experienceUnderstand identity verification, KYC and onboarding data as a data domain, not just as a product surfacePartner with Risk, Compliance, and Legal on regulatory obligations tied to fraud and identity controlsOwn incident response leadership when major fraud events, model failures, or data integrity failures occur"
Note there are missing spaces but we keep as is.
Then "Bonus points" bullet list:
- Have worked at a scaling startup previously.
- Have experience with data mesh or domain-oriented data ownership models in practice, not just in theory.
- Have experience working with stablecoin and cryptocurrencies.
- Have experience working with AI in a production environment.
- Have worked in environments with k8s / gRPC / Spring Boot / Java / Python.
Then "AI Literacy & Tools" paragraph:
"As Zepz, we're building AI into how we work. We're on a journey to making AI utilisation streamline what we do. "For this position we are looking for:Proficient depth of knowledge around automation, AI and how those can be applied in your area of work — from accelerating analysis and insight generation, to improving how we design and deliver processes and communications.Practical experience using AI tools in a professional context — whether to connect information, draft content, build frameworks, or create insights from data. We want people who actively use AI to increase productivity, not those who are waiting to be told to.Strong AI literacy — an understanding of what AI assistants can and cannot do, how to prompt effectively, and how to critically evaluate AI-generated output before using it.Awareness of the ethical considerations and responsible use of AI in the workplace, particularly in the context of sensitive data, customer experience, and compliance.Prior experience using Claude (Anthropic) is a nice-to-have but not essential — what matters most is that you are genuinely engaged with AI as a tool, comfortable experimenting, and eager to share what works with your wider team"
Then final paragraph about applying etc (boilerplate) - we remove.
Now we need to decide sections.
We have About the role (prose). Responsibilities: the "What you will own" list (including sublist). Could also include "In your first few months, you will:" as part of responsibilities.
Requirements/Qualifications: Could combine "What you bring to the table", "FinCrime & Identity Domain", "Bonus points", "AI Literacy & Tools" as qualifications/skills. But we need to decide headings: maybe "Requirements" for mandatory, "Preferred Qualifications" for bonus points, but we only have allowed headings: About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule. We can use Requirements and Qualifications and Skills as separate if needed. But we must not create sections with filler. We have content for Requirements (mandatory), Qualifications (preferred), Skills (maybe AI literacy). However we could combine all into one "Requirements" section if we want fewer sections. The instruction: "Prefer a few substantive sections over many thin ones." So we could combine all qualification-like content into a single "Requirements" section (or "Qualifications") and not create multiple thin sections. Let's decide to use "Requirements" for all mandatory and preferred qualifications, and maybe "Skills" for AI literacy? But that would be thin. Better to put all under "Requirements". However we need to ensure we don't lose nuance; but we can keep all bullet points under Requirements.
Thus sections: About the role (prose), Responsibilities (list), Requirements (list of all qualification bullets plus domain, bonus, AI literacy), Benefits (list). That's four sections.
Now we need to format.
About the role:
About the role
... the prose ...
Responsibilities: Responsibilities
- ... each bullet ...
We need to include the sub-bullets under "In your first few months, you will:" as separate - items.
We must preserve all bullet points
for prose. We must not repeat job title as heading. No Job ID. We must preserve all substantive information; do not invent or summarize away detail. Must keep details. We need to extract content precisely. Let's parse: About the role paragraph: Starting after "About the role". The text: "This is a high-visibility, high-accountability role — outcomes are reported to C-level and the board, and the role owns both the data strategy and the model performance behind systems that directly protect revenue and customer trust. The mandate in one line: own what data is generated and how it is generated and consumed across the Identity and Financial Crime domains. Almost every consequential decision Zepz makes about a transaction — is this person who they say they are, is this payment safe, is this account compromised — is a decision made on data. Identity is our single largest producer of that data. Financial Crime is our single largest consumer of it. Both sit in the same engineering organization, and today no one owns the path between them. The result is a permanent tax on everything downstream: problems discovered late, by the people least able to fix them at source. This role exists to remove that tax. You will lead a team of 25+ Data Scientists, Machine Learning Engineers and Analytics Engineers across the Identity and Financial Crime domains" That's the About the role prose. Then "What you will own" bullet list: - Own the definition, generation, quality and consumption of data across the Identity and Financial Crime domains — what gets produced, what it means, what guarantees come with it, and who is accountable when those guarantees are not met. - Make data quality an engineering discipline rather than a downstream cleanup exercise: producer/consumer contracts, freshness and integrity monitoring, lineage, a named owner for every critical dataset, and a real path from a detected anomaly to a fix at source. You'll be measured on whether downstream teams stop discovering data problems by accident. - Own the models and analytics behind how Zepz detects fraud and financial crime — feature engineering, model development, deployment, monitoring and retraining. You own model performance in production, not just model development. - Lead a mixed organization of Data Scientists, Machine Learning Engineers and Analytics Engineers. These are three different crafts with three different career paths, and leading them well means understanding all three rather than favoring the one you came from. - Partner as a peer with the Engineering Manager who owns the fraud detection platform and reports directly to the VP Engineering. Your models run on their platform: you own model outcomes, they own platform outcomes, and you jointly own the seam between them. - Partner with our central data function, which owns core data infrastructure and platform tooling. You are a demanding customer and a collaborator, not a replacement — you build domain data products and quality practice on top of what they provide, and feed real requirements back to them. - Set standards with Identity engineering teams who do not report to you. Much of your impact depends on changing how data is produced upstream, which is influence work: making the cost of poor data visible with evidence rather than assertion, and making it easy for those teams to do the right thing. - Create high-performing teams with a bias toward shipping without sacrificing reliability or trust in the numbers. Use technical and data debt as a tool to move faster, while ensuring your teams have the space to build it right once they know what the right thing to build is. - Hire, retain, and level up superlative talent to ensure our team remains world-class as we continue to scale and launch into new markets. - Roll up your sleeves when models degrade or data breaks. You'll need to discuss technical solutions, review work in depth, and help execute strategy with your leads. - In your first few months, you will: - Establish an honest, evidenced baseline of data quality across the identity-to-financial-crime path — what breaks, how often, who finds out, and what it costs in downstream hours and model degradation. Bring the number, not the anecdote. - Agree a written ownership model with the central data team and the Identity engineering leads: who owns which datasets, what the contracts are, and what happens when one is broken. - Instrument the critical path, so data issues are detected by monitoring rather than by an analyst noticing that something looks odd. - Get to know the models — how they perform, how they degrade, how retraining actually works today, and where risk is concentrated. - Assess the team: identify the gaps, the flight risks and the people ready for more, and start hiring against a plan rather than against a vacancy. - Promote a culture of ownership, accountability, and shipping. Note: The bullet list includes a sublist under "In your first few months, you will:" which is also bullet points. We'll treat as part of responsibilities. Now "What you bring to the table" bullet list: - You have a track record of leading a data organization of comparable size (roughly 15–25 people) and of supporting the professional development of managers and senior-level ICs. - Track record of leading mixed technical teams — Data Scientists, ML Engineers, Analytics Engineers — with genuine credibility across all three crafts. - Demonstrable ownership of data quality at scale: contracts, observability, lineage, semantic consistency — with evidence of the downstream impact, not just the tooling you installed. - Fluency with modern lakehouse and analytics engineering practice: Databricks (or equivalent), dbt-style transformation, feature store concepts and real-time feature serving, and the discipline of treating datasets as products with owners and SLAs. - Experience with production ML in a real-time context: model monitoring, drift detection, and retraining pipelines. - Proven experience with AWS at scale. - Direct experience leading FinCrime/Fraud or Identity engineering or data science teams (not just adjacent risk/security). - The ability to get outcomes from teams that don't report to you. We'd rather hear about a time you changed a producer team's behavior than a time you escalated. - At least 8+ years of technical experience in a hybrid IC/management role. - You can attract and retain top talent and lead teams with diverse skill sets. - You work well with Engineering and Product leaders and are comfortable building effective relationships with folks outside of the Prod-Eng spheres. - You have exceptional judgment; we're scaling quickly — you'll be required to make good tradeoffs. Then "FinCrime & Identity Domain" paragraph: "Bring prior experience leading FinCrime/Fraud or Identity teams — understanding of typologies (card fraud, ATO, first-party fraud, synthetic identity, money mule networks), the regulatory context, and how detection strategy trades off against customer experienceUnderstand identity verification, KYC and onboarding data as a data domain, not just as a product surfacePartner with Risk, Compliance, and Legal on regulatory obligations tied to fraud and identity controlsOwn incident response leadership when major fraud events, model failures, or data integrity failures occur" Note there are missing spaces but we keep as is. Then "Bonus points" bullet list: - Have worked at a scaling startup previously. - Have experience with data mesh or domain-oriented data ownership models in practice, not just in theory. - Have experience working with stablecoin and cryptocurrencies. - Have experience working with AI in a production environment. - Have worked in environments with k8s / gRPC / Spring Boot / Java / Python. Then "AI Literacy & Tools" paragraph: "As Zepz, we're building AI into how we work. We're on a journey to making AI utilisation streamline what we do. "For this position we are looking for:Proficient depth of knowledge around automation, AI and how those can be applied in your area of work — from accelerating analysis and insight generation, to improving how we design and deliver processes and communications.Practical experience using AI tools in a professional context — whether to connect information, draft content, build frameworks, or create insights from data. We want people who actively use AI to increase productivity, not those who are waiting to be told to.Strong AI literacy — an understanding of what AI assistants can and cannot do, how to prompt effectively, and how to critically evaluate AI-generated output before using it.Awareness of the ethical considerations and responsible use of AI in the workplace, particularly in the context of sensitive data, customer experience, and compliance.Prior experience using Claude (Anthropic) is a nice-to-have but not essential — what matters most is that you are genuinely engaged with AI as a tool, comfortable experimenting, and eager to share what works with your wider team" Then final paragraph about applying etc (boilerplate) - we remove. Now we need to decide sections. We have About the role (prose). Responsibilities: the "What you will own" list (including sublist). Could also include "In your first few months, you will:" as part of responsibilities. Requirements/Qualifications: Could combine "What you bring to the table", "FinCrime & Identity Domain", "Bonus points", "AI Literacy & Tools" as qualifications/skills. But we need to decide headings: maybe "Requirements" for mandatory, "Preferred Qualifications" for bonus points, but we only have allowed headings: About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule. We can use Requirements and Qualifications and Skills as separate if needed. But we must not create sections with filler. We have content for Requirements (mandatory), Qualifications (preferred), Skills (maybe AI literacy). However we could combine all into one "Requirements" section if we want fewer sections. The instruction: "Prefer a few substantive sections over many thin ones." So we could combine all qualification-like content into a single "Requirements" section (or "Qualifications") and not create multiple thin sections. Let's decide to use "Requirements" for all mandatory and preferred qualifications, and maybe "Skills" for AI literacy? But that would be thin. Better to put all under "Requirements". However we need to ensure we don't lose nuance; but we can keep all bullet points under Requirements. Thus sections: About the role (prose), Responsibilities (list), Requirements (list of all qualification bullets plus domain, bonus, AI literacy), Benefits (list). That's four sections. Now we need to format. About the role:
About the role
... the prose ...
Responsibilities:Responsibilities
- ... each bullet ...