Senior Product Manager, AI Data Platform
for section headings (e.g. About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule)." So we can have About the role section containing the description of the role and maybe company overview.
We need to avoid repeating job title as heading. No Job ID.
We need to remove boilerplate: cookie/navigation/legal/EEO/equal-opportunity/scam-warning boilerplate, duplicated lines, application instructions. So we should remove the sections: "Our Commitment to Diversity, Equity & Inclusion", "Equal Employment Opportunity Statement", "This job description is intended..." etc. Also "Please note: MGT does not accept unsolicited resumes..." is application instruction; remove. Also "Compensation $190,000 - $300,000 annual base salary, plus bonus and equity." That's Pay section.
We need to extract responsibilities: "What You Own" bullet points. Also maybe "What Great Looks Like" could be responsibilities or qualifications? It's more like expectations; could be part of Responsibilities or Qualifications. We'll treat as Responsibilities.
Requirements: "Minimum Requirements". Also "Ideal Candidates" maybe qualifications/skills.
Skills: maybe list of skills like evaluation harnesses, data/API vendor stack, engineering background, curiosity about AI search, growth stage experience, references.
Benefits: Not mentioned besides bonus and equity (that's compensation). No explicit benefits like health, 401k. So we should not create Benefits section unless posting actually states them. It does not, so omit Benefits.
Schedule: Not mentioned. Omit.
Pay: we have compensation range. So include Pay section.
Thus sections: About the role, Responsibilities, Requirements, Qualifications, Skills, Pay.
We need to use - for any list of duties/requirements/benefits. For prose use
.
We need to preserve all substantive information; do not invent or summarize away detail.
We need to avoid filler like "commensurate with experience". Not present.
Let's extract content.
About the role: combine "About MGT Insurance" paragraph and maybe the role description: "We're hiring a Senior Product Manager to wake up every day thinking about driving quality results and expanding our datasets in tight coordination with the Insurance product team. You own and deliver the vendor stack, the quality and hit rate, and telemetry and evals to prove it. This is not a roadmap-and-jira role. Product and engineering lanes are merging at MGT, and product talent operates fluidly across strategy, data, design, and engineering. PMs that succeed at MGT embrace end-to-end execution and lean into our comparative advantage: agility. You will write compelling PRDs and then execute, directly contributing code, analyzing data in our warehouse, and own results from concept through to post go live."
We'll put that in
maybe multiple paragraphs.
Responsibilities: from "What You Own" bullet list:
- Manage our data and AI web search vendors as a portfolio: cost per successful enrichment, hit rate, field-level quality, latency, and contract terms.
- Onboard new vendors, run bake-offs with real evals, and kill or renegotiate underperformers on the numbers.
- Quality and telemetry at scale. Instrument every vendor call end to end. Build (yourself, in Python/TypeScript) an agent that monitors hit rate, drift, outages, and quality regressions across the pipeline and flags them before ops or underwriting notice.
- The frontier of AI search and extraction. Stay current on AI web search and retrieval options (native model search tools, search APIs, crawlers, structured extraction) and know which one wins for which entity type and why, with evals to back it.
- The enrichment roadmap. Decide which new signals move loss ratio, bind rate, or underwriting speed, and sequence them. Make crisp build-vs-rent calls and defend the unit economics.
Also "What Great Looks Like" could be responsibilities or qualifications; but it's more like expected outcomes. Could be included under Responsibilities as well. Let's treat as Responsibilities bullet list:
- Every vendor is measured with hit rate, quality, latency, and cost per successful enrichment, so that you can leverage those metrics to make renewal decisions.
- A monitoring agent you built catches vendor degradation and pipeline drift automatically; incidents are found by the system, not by our underwriters.
- You can name the best AI search option for each of our core entity types and show the eval results that prove it.
- Enrichment quality improvements you shipped are traceable to a measurable change in underwriting speed or accuracy.
We'll include these as responsibilities.
Requirements: "Minimum Requirements":
- You have shipped meaningful work (beyond simple prototypes) with an AI engineering tool (Claude Code, Codex, Devin, etc.) and can show the commits.
- You have 5+ years of experience at a firm that achieved scale, ideally InsureTech, FinTech, or Consulting.
- You build things directly. Nobody is on an island at MGT, but everyone ships.
We'll put as list.
Qualifications: maybe "Ideal Candidates" bullet list:
- Evals are your native language. You've built evaluation harnesses for LLM or retrieval systems and can talk about precision, recall, and cost tradeoffs.
- You've run a data or API vendor stack. You have owned third-party data integrations, measured their quality, and made the call to add or remove one.
- Engineering background. You wrote code the old fashioned way pre-AI, and are using the latest tools to increase your impact.
- Curiosity about the AI search landscape. You have opinions, backed by experiments, on the current AI web search and extraction options.
- Growth stage experience. MGT is a small, high-accountability organization and that's a plus for you.
- References from peer department managers (Engineering, Data, Underwriting) that loved working with you.
We'll treat as Qualifications.
Skills: Could be overlapping with qualifications; but we can extract skill set: evaluation harnesses, precision/recall/cost tradeoffs, data/API vendor stack, third-party data integrations, Python/TypeScript, AI engineering tools, etc. However we already have qualifications list; maybe we can skip Skills section to avoid duplication. But instruction: prefer a few substantive sections over many thin ones. We could combine Qualifications and Skills? But they ask to use those headings if content exists. We have content for Qualifications; we could also have Skills if we can extract distinct skill list not already covered. However the Ideal Candidates list already covers skills. To avoid duplication, we could omit Skills section. But we need to include sections only when posting has real specific content for it. We have content for Skills? The posting does not have a separate "Skills" heading; but we could infer skills from Ideal Candidates. However we already used Ideal Candidates for Qualifications. Might be okay to not have Skills section.
Pay: compensation range: "$190,000 - $300,000 annual base salary, plus bonus and equity."
We'll put in
.
Now produce HTML fragment.
Structure:
About the role
...
Responsibilities
- ...
...
Requirements
- ...
...
Qualifications
- ...
...
Pay
$190,000 - $300,000 annual base salary, plus bonus and equity.
Make sure no extra sections.
We need to preserve all substantive information; ensure we didn't lose any.
Let's craft the About the role paragraph(s). Could split into two paragraphs: company description and role description.
Company description: "MGT is the first AI-driven, neo-insurer focused on evolving commercial P&C insurance for the brokers and small business owners. By combining the expertise of industry veterans with state-of-the-art technology, we are ushering in the next evolution of commercial insurance and working to move the industry forward through the use of modern technology, better processes, and a rock-star team. We seek visionary leaders who thrive in dynamic, entrepreneurial settings, and excel in autonomous roles. MGT’s AI Data Platform is the engine for our Agentic Underwriting: the vendor data, AI web search, extraction, and scoring that runs behind every quote MGT produces."
Role description: "We're hiring a Senior Product Manager to wake up every day thinking about driving quality results and expanding our datasets in tight coordination with the Insurance product team. You own and deliver the vendor stack, the quality and hit rate, and telemetry and evals to prove it. This is not a roadmap-and-jira role. Product and engineering lanes are merging at MGT, and product talent operates fluidly across strategy, data, design, and engineering. PMs that succeed at MGT embrace end-to-end execution and lean into our comparative advantage: agility. You will write compelling PRDs and then execute, directly contributing code, analyzing data in our warehouse, and own results from concept through to post go live."
We'll put both as under About the role.
Now Responsibilities list: combine What You Own and What Great Looks Like.
Let's list each bullet as
- .
Make sure to preserve details exactly.
What You Own bullets:
1. Manage our data and AI web search vendors as a portfolio: cost per successful enrichment, hit rate, field-level quality, latency, and contract terms.
2. Onboard new vendors, run bake-offs with real evals, and kill or renegotiate underperformers on the numbers.
3. Quality and telemetry at scale. Instrument every vendor call end to end. Build (yourself, in Python/TypeScript) an agent that monitors hit rate, drift, outages, and quality regressions across the pipeline and flags them before ops or underwriting notice.
4. The frontier of AI search and extraction. Stay current on AI web search and retrieval options (native model search tools, search APIs, crawlers, structured extraction) and know which one wins for which entity type and why, with evals to back it.
5. The enrichment roadmap. Decide which new signals move loss ratio, bind rate, or underwriting speed, and sequence them. Make crisp build-vs-rent calls and defend the unit economics.
What Great Looks Like bullets:
1. Every vendor is measured with hit rate, quality, latency, and cost per successful enrichment, so that you can leverage those metrics to make renewal decisions.
2. A monitoring agent you built catches vendor degradation and pipeline drift automatically; incidents are found by the system, not by our underwriters.
3. You can name the best AI search option for each of our core entity types and show the eval results that prove it.
4. Enrichment quality improvements you shipped are traceable to a measurable change in underwriting speed or accuracy.
We'll include all.
Requirements list:
- You have shipped meaningful work (beyond simple prototypes) with an AI engineering tool (Claude Code, Codex, Devin, etc.) and can show the commits.
- You have 5+ years of experience at a firm that achieved scale, ideally InsureTech, FinTech, or Consulting.
- You build things directly. Nobody is on an island at MGT, but everyone ships.
Qualifications list (Ideal Candidates):
- Evals are your native language. You've built evaluation harnesses for LLM or retrieval systems and can talk about precision, recall, and cost tradeoffs.
- You've run a data or API vendor stack. You have owned third-party data integrations,
. We need to preserve all substantive information; do not invent or summarize away detail. We need to avoid filler like "commensurate with experience". Not present. Let's extract content. About the role: combine "About MGT Insurance" paragraph and maybe the role description: "We're hiring a Senior Product Manager to wake up every day thinking about driving quality results and expanding our datasets in tight coordination with the Insurance product team. You own and deliver the vendor stack, the quality and hit rate, and telemetry and evals to prove it. This is not a roadmap-and-jira role. Product and engineering lanes are merging at MGT, and product talent operates fluidly across strategy, data, design, and engineering. PMs that succeed at MGT embrace end-to-end execution and lean into our comparative advantage: agility. You will write compelling PRDs and then execute, directly contributing code, analyzing data in our warehouse, and own results from concept through to post go live." We'll put that in
maybe multiple paragraphs. Responsibilities: from "What You Own" bullet list: - Manage our data and AI web search vendors as a portfolio: cost per successful enrichment, hit rate, field-level quality, latency, and contract terms. - Onboard new vendors, run bake-offs with real evals, and kill or renegotiate underperformers on the numbers. - Quality and telemetry at scale. Instrument every vendor call end to end. Build (yourself, in Python/TypeScript) an agent that monitors hit rate, drift, outages, and quality regressions across the pipeline and flags them before ops or underwriting notice. - The frontier of AI search and extraction. Stay current on AI web search and retrieval options (native model search tools, search APIs, crawlers, structured extraction) and know which one wins for which entity type and why, with evals to back it. - The enrichment roadmap. Decide which new signals move loss ratio, bind rate, or underwriting speed, and sequence them. Make crisp build-vs-rent calls and defend the unit economics. Also "What Great Looks Like" could be responsibilities or qualifications; but it's more like expected outcomes. Could be included under Responsibilities as well. Let's treat as Responsibilities bullet list: - Every vendor is measured with hit rate, quality, latency, and cost per successful enrichment, so that you can leverage those metrics to make renewal decisions. - A monitoring agent you built catches vendor degradation and pipeline drift automatically; incidents are found by the system, not by our underwriters. - You can name the best AI search option for each of our core entity types and show the eval results that prove it. - Enrichment quality improvements you shipped are traceable to a measurable change in underwriting speed or accuracy. We'll include these as responsibilities. Requirements: "Minimum Requirements": - You have shipped meaningful work (beyond simple prototypes) with an AI engineering tool (Claude Code, Codex, Devin, etc.) and can show the commits. - You have 5+ years of experience at a firm that achieved scale, ideally InsureTech, FinTech, or Consulting. - You build things directly. Nobody is on an island at MGT, but everyone ships. We'll put as list. Qualifications: maybe "Ideal Candidates" bullet list: - Evals are your native language. You've built evaluation harnesses for LLM or retrieval systems and can talk about precision, recall, and cost tradeoffs. - You've run a data or API vendor stack. You have owned third-party data integrations, measured their quality, and made the call to add or remove one. - Engineering background. You wrote code the old fashioned way pre-AI, and are using the latest tools to increase your impact. - Curiosity about the AI search landscape. You have opinions, backed by experiments, on the current AI web search and extraction options. - Growth stage experience. MGT is a small, high-accountability organization and that's a plus for you. - References from peer department managers (Engineering, Data, Underwriting) that loved working with you. We'll treat as Qualifications. Skills: Could be overlapping with qualifications; but we can extract skill set: evaluation harnesses, precision/recall/cost tradeoffs, data/API vendor stack, third-party data integrations, Python/TypeScript, AI engineering tools, etc. However we already have qualifications list; maybe we can skip Skills section to avoid duplication. But instruction: prefer a few substantive sections over many thin ones. We could combine Qualifications and Skills? But they ask to use those headings if content exists. We have content for Qualifications; we could also have Skills if we can extract distinct skill list not already covered. However the Ideal Candidates list already covers skills. To avoid duplication, we could omit Skills section. But we need to include sections only when posting has real specific content for it. We have content for Skills? The posting does not have a separate "Skills" heading; but we could infer skills from Ideal Candidates. However we already used Ideal Candidates for Qualifications. Might be okay to not have Skills section. Pay: compensation range: "$190,000 - $300,000 annual base salary, plus bonus and equity." We'll put in
. Now produce HTML fragment. Structure:
About the role
...
Responsibilities
- ... ...
Requirements
- ... ...
Qualifications
- ... ...
Pay
$190,000 - $300,000 annual base salary, plus bonus and equity.
Make sure no extra sections. We need to preserve all substantive information; ensure we didn't lose any. Let's craft the About the role paragraph(s). Could split into two paragraphs: company description and role description. Company description: "MGT is the first AI-driven, neo-insurer focused on evolving commercial P&C insurance for the brokers and small business owners. By combining the expertise of industry veterans with state-of-the-art technology, we are ushering in the next evolution of commercial insurance and working to move the industry forward through the use of modern technology, better processes, and a rock-star team. We seek visionary leaders who thrive in dynamic, entrepreneurial settings, and excel in autonomous roles. MGT’s AI Data Platform is the engine for our Agentic Underwriting: the vendor data, AI web search, extraction, and scoring that runs behind every quote MGT produces." Role description: "We're hiring a Senior Product Manager to wake up every day thinking about driving quality results and expanding our datasets in tight coordination with the Insurance product team. You own and deliver the vendor stack, the quality and hit rate, and telemetry and evals to prove it. This is not a roadmap-and-jira role. Product and engineering lanes are merging at MGT, and product talent operates fluidly across strategy, data, design, and engineering. PMs that succeed at MGT embrace end-to-end execution and lean into our comparative advantage: agility. You will write compelling PRDs and then execute, directly contributing code, analyzing data in our warehouse, and own results from concept through to post go live." We'll put both asunder About the role. Now Responsibilities list: combine What You Own and What Great Looks Like. Let's list each bullet as