Data Scientist – Core Machine Learning // San, Jose, CA// In- Person Interview Must

Data Scientist – Core Machine Learning

In- Person Interview Must

San Jose, CA

Feature Engineering | Model Architecture | End-to-End Pipelines | Model Routing & Observability

Position Summary

Hands-on Data Scientist to design, build and productionize predictive and recommendation models on large-scale enterprise data. This is a core data science role, not a GenAI application development role. GenAI familiarity is welcome, but profiles centred on prompt engineering, RAG assembly or chatbot/agent development without concurrent classical or deep ML model-building ownership are not a fit.

Key Responsibilities

• Problem formulation & features — Translate ambiguous business problems into well-posed ML problems; run EDA and statistical validation; design predictive features (aggregations, embeddings, temporal and behavioural signals); own data quality, leakage prevention, sampling and train/validation/test design.
• Model architecture — Select and architect the right model family (regression/classification, gradient-boosted trees, clustering, time-series and anomaly detection, deep learning, recommendation/ranking); design from first principles rather than AutoML; justify trade-offs on accuracy, latency, interpretability and cost.
• Experimentation — Hyperparameter optimization, cross-validation, ablation and error analysis; define and defend metrics, baselines and statistical significance for A/B and online tests.
• End-to-end pipelines — Own ingestion → feature computation → training → deployment → monitoring; automate retraining, versioning and lineage (MLflow or equivalent); deploy scalable model services instrumented for drift, degradation and rollback.
• Model routing & selection logic — Where multiple models or agents serve a workflow, define routing rules and the validation logic applied when switching models; use semantic/embedding-based mapping of task or agent goals to the appropriate model, with confidence thresholds, fallback order and post-switch output validation.
• Observability & autonomous correction — Instrument models and pipelines for runtime observability; implement auto-correct and auto-trigger rules (retry, re-route, auto-start of an alternate model or agent, kill-switch) so that failures such as timeouts, low-confidence outputs or degraded responses are handled without human intervention.
• Drift detection & alerting — Define data and concept drift metrics with explicit thresholds; build automated alerting, kill-switch and escalation paths, and trigger retraining or rollback when thresholds are breached as upstream data changes.
• Collaboration — Work in agile cross-functional squads; explain model behaviour, limitations and business impact at whiteboard level to technical and non-technical stakeholders; mentor junior data scientists.

Required Qualifications

• Bachelor’s or Master’s in CS, Statistics, Mathematics or a related quantitative discipline; PhD a plus.
• 8+ years applied ML / data science with a verifiable record of models sustained in production.
• Strong mathematical foundations — linear algebra, probability, statistics, optimization.
• Demonstrated depth in feature engineering and model architecture design, explainable on a whiteboard for models personally built.
• Proven end-to-end pipeline ownership, not notebook-stage experimentation.
• Expert Python and SQL; pandas, NumPy, scikit-learn and PyTorch or TensorFlow.
• Distributed processing (Spark/PySpark, Hadoop, BigQuery) and a cloud ML platform (SageMaker, Azure ML or Vertex AI).
• MLOps working knowledge — MLflow, Docker, Kubernetes, CI/CD, model monitoring.
• Production model observability and drift management — threshold definition, automated alerting, kill-switches and retraining/rollback triggers.
• Experience designing model routing or fallback logic across multiple models, including the validation applied when a switch occurs.

Preferred

• Recommendation, ranking or personalization engines at scale; time-series forecasting and anomaly detection on telemetry data.
• NLP with classical and transformer-based approaches (embeddings, NER, classification, topic modelling).
• LLM/GenAI exposure as a complement to core ML depth; enterprise technology, networking or telecom domain experience; Tableau or Power BI.

Screening Focus

Interview Process

• Internal technical screen on core ML depth.
• Client whiteboard discussion on feature engineering, model architecture and pipeline design.
• Client hiring manager discussion on delivery experience and stakeholder communication.