Storyful is an equal opportunity employer
Job Description
Job Title : Lead Data Scientist
Location: Dublin (Hybrid — 3 days/week in office)
Reporting to: CPTO (Storyful)
About Storyful
Storyful (a News Corp company) helps organisations discover, verify, and act on information at speed — from breaking news to complex reputation and narrative risk.
We're building RiskRadar: an AI-native, decision-ready Narrative & Reputational Risk platform that fuses signals from licensed content, social, and broadcast/video into explainable scoring and clear next actions for high-scrutiny brands — including financial services buyers with serious governance expectations.
This is a pivotal hire: the first Data Scientist at Storyful. You'll set the bar for how we build trustworthy ML/AI that ships.
Role Mission
Build and production is the core data science capabilities behind RiskRadar — specifically:
* a robust, explainable Reputation / Narrative Risk scoring system, and
* entity resolution across messy, multi-source datasets (brands, products, execs, organisations, locations, subsidiaries).
Bonus (strategic): help us explore vision + multimodal models to detect manipulated / misleading video for our Newswire verification workflows.
What You'll Do (Hands-on Responsibilities)
* Reputation & Narrative Risk Scoring (Explainable + Governed)
* Design a score framework (overall + sub-scores) that is measurable over time, resistant to gaming, and useful in executive decision-making.
* Build scoring that can answer: "What changed, why, and what evidence supports this?"
* Implement calibration, confidence, and uncertainty so customers understand when to trust the model and when to escalate to humans.
* Set up model governance patterns aligned to model risk management expectations (audit trails, versioning, monitoring, documentation, change control).
* Entity Resolution (Truth Layer Across Sources)
* Build entity resolution pipelines that unify entities across licensed news, social, and broadcast/video metadata.
* Combine the right approaches: rules where they win, probabilistic matching where it matters, embeddings/LLM-assisted linking where it scales — always with measurable quality.
* Establish golden datasets, error taxonomies, and repeatable evaluation so entity resolution improves continuously.
* Ship ML Like a SaaS Operator (Fast Experiments → Production)
* Run crisp experiments: hypotheses, baselines, metrics, iteration loops — and kill weak ideas early.
* Partner with Product + Engineering to take ML/AI features to production quickly, safely, and cost-effectively.
* Build evaluation harnesses for ML and LLM components (offline + human review + online measurement).
* Implement production standards: monitoring, drift detection, cost/latency controls, incident playbooks, and quality dashboards.
* Founding DS Leadership (Today: IC; Soon: Build the Team)
* Establish best practices for experimentation, reproducibility, documentation, and responsible AI.
* Lead cross-functional delivery with Product, Engineering, and AI teams across News Corp.
* Within 12 months: help hire/mentor 1–2 additional DS/ML roles (scope dependent), while staying hands-on.
* Bonus: Video Verification ML (Vision / Multimodal)
* Prototype approaches for manipulated media detection and video authenticity signals.
* Translate "model output" into a verification workflow with clear confidence and evidence, not black-box answers.
Tech & Working Environment
* Cloud: AWS
* LLM stack: LangChain (or equivalent patterns), Langfuse for tracing/observability, modern LLM APIs
* Core: Python, SQL, data pipelines, model packaging + CI/CD
* You'll work closely with our AI Architect, Product, Engineering, and verification experts.
What You'll Bring (Requirements)
Must-have
* 7+ years in applied Data Science / ML, with multiple production deployments in a commercial environment (SaaS strongly preferred).
* Proven experience leading cross-functional teams to deliver production-ready ML/AI (even if you weren't the people manager).
* Strong grounding in: classification/scoring/ranking, NLP (and/or LLM applications), statistics, evaluation, and experimentation.
* Demonstrated ability to build explainable systems: not just performance, but transparency, evidence, and user trust.
* Experience designing evaluation strategies: labeled datasets, human-in-the-loop review, acceptance thresholds, monitoring and drift.
* Comfortable operating in ambiguity with high ownership and high pace.
Strong advantage
* Deep experience with entity resolution / record linkage at scale (probabilistic matching, embeddings, graph-based approaches).
* Experience building for regulated / high-governance contexts (financial services is a plus): auditability, documentation, controls.
* LLM evaluation and reliability methods (prompt eval, retrieval eval, hallucination mitigation, guardrails).
* Computer vision / multimodal experience, especially around authenticity, manipulation detection, or media forensics.
What Success Looks Like
First 90 days
* Clear score and entity resolution strategy, baselines, datasets, and metrics agreed with Product/Engineering.
* First production-ready scoring and entity resolution increments shipped behind feature flags.
* Evaluation + monitoring foundations in place (including Langfuse tracing standards for LLM workflows).
By 6 months
* Explainable scoring system with evidence trails, confidence, and drift monitoring live for real users.
* Entity resolution quality improving on a measurable cadence (golden set + error reduction plan).
* Governance pack in place suitable for financial services buyers (documentation, audit trails, change controls).
By 12 months
* Mature experimentation-to-production loop: faster iteration, lower incident rate, clearer model performance visibility.
* You've begun mentoring/hiring to expand DS capability while remaining a hands-on technical leader.
* Optional: early vision/multimodal verification prototypes validated with newsroom workflows.
Why This Role is Different
* You're not inheriting a mature DS org — you're founding it.
* You'll build AI that executives will rely on under pressure — where explainability and governance aren't "nice to have", they're the product.
* You'll ship. A lot. And you'll help define what "good" looks like for Storyful's AI future.
Job Category
Storyful - Product & Technology