About the role
Role Summary
We are looking for a hands-on Data Scientist with a strong statistics and AI foundation to work across raw, unstructured, and behavioral data — engineering features, building advanced models, and shipping production-grade deliverables in close collaboration with the client's data and engineering teams.
Key Responsibilities
- Mine raw credit bureau (tradeline-level) and bank transaction data to identify, structure, and validate discrete, machine-readable risk signals with clear trigger conditions and supporting evidence.
- Apply LLM-based extraction techniques to convert unstructured data (e.g., sales call transcripts) into structured, schema-validated signals (claim credibility, objection type/resolution, sentiment) suitable for downstream models — not narrative summaries.
- Engineer features from high-frequency, noisy behavioral/interaction event data using sequence and trajectory analysis, clustering, and state-based modeling.
- Build and validate advanced models spanning classical ML, survival/hazard modeling (Cox, parametric, ML survival models), and Bayesian methods (hierarchical models, empirical Bayes, sequential updating).
- Rigorously test whether new signals/features add incremental predictive value beyond existing structured models — avoid replicating what current models already capture.
- Take models and features from experimentation through to production-grade deliverables: version-controlled code, reproducible pipelines, and monitored production services.
- Own Python codebases with GitHub-based version control and CI/CD workflows for feature/model deployment.
- Contribute reusable content, frameworks, and learnings to Affine's internal AI Center of Excellence (CoE).
- Operate entirely within access-controlled infrastructure, working only with de-identified data under named-user access.