Job highlights

- Expertise in problem framing for ML, robust applied statistics and experimentation skills, solid supervised ML fundamentals, advanced SQL and data modeling proficiency, and Python programming for reproducible workflows
- Translate business questions into ML problems, design experiments, develop and validate models, manage end-to-end project delivery, and communicate results to stakeholders

Mandatory 1) Problem framing + product sense

- Translate ambiguous business questions into measurable ML/data problems.
- Define success metrics, constraints (latency, cost, risk), and an evaluation plan.
- Decide when not to use ML and propose simpler alternatives.

2) Strong statistics + experimentation

- Hypothesis testing, confidence intervals, power/sample size, bias/variance.
- A/B testing design and pitfalls (multiple testing, peeking, novelty effects).
- Causal thinking basics (confounding, selection bias); at least one causal method in practice (e.g., matching, diff-in-diff, IV, uplift).

3) Solid ML fundamentals (applied)

- Supervised learning: linear/logistic regression, tree-based models, boosting, regularization, calibration.
- Model evaluation: proper metrics, cross-validation, leakage prevention, error analysis.
- Feature engineering, imbalanced learning, interpretability techniques.

4) Data fluency: SQL + data modeling

- Write complex SQL (joins, windows, CTEs), debug data issues, validate pipelines.
- Understand data schemas, grain, dimensional modeling basics, data quality checks.

5) Programming + reproducible workflows

- Proficiency in Python (or R) for data/ML work; ability to write production-quality notebooks/scripts.
- Code quality: modularity, testing basics, version control (Git), packaging/environment management.
- Clear documentation and reproducible analyses.

6)



Communication + stakeholder management

- Explain tradeoffs and results to non-technical audiences; tell a coherent data story.
- Write concise decision docs; align on metrics and scope; handle pushback.

7) End-to-end delivery ownership

- Take a project from data discovery modeling validation launch plan monitoring.
- Identify risks early (data availability, privacy, feedback loops) and drive mitigation.

Good to have 1) Production ML / MLOps depth

- Model serving patterns (batch vs real-time), CI/CD, feature stores, model registry.
- Monitoring: data drift, performance decay, alerting; retraining strategies.
- Familiarity with Docker/Kubernetes and cloud ML stacks (AWS/GCP/Azure).

2) Deep learning (when relevant)

- Practical experience with PyTorch/TensorFlow, embeddings, transformers, training/inference tradeoffs.
- Domain-specific DL (NLP, CV, recommender systems).

3) Data engineering / big data tooling

- Spark/Databricks, distributed computing, orchestration (Airflow), streaming (Kafka).
- Building reliable datasets, backfills, incremental loads.

4) Advanced causal inference

- Causal graphs (DAGs), doubly robust estimators, synthetic controls, mediation, Bayesian causal.
- Designing quasi-experiments in messy real-world settings.

5) Optimization + decision science

- Linear/integer programming, bandits, reinforcement learning (select use cases).
- Forecasting at scale (hierarchical, probabilistic).

6) Domain expertise

- E.g., fraud/risk, ads/auction systems, supply chain, pricing, healthcare, fintech compliance.
- Knowing domain constraints often matters as much as algorithm choice.

7) Leadership beyond your own work

- Mentoring, setting best practices, reviewing others analyses/models.
- Influencing roadmap, creating reusable frameworks, raising the team’s bar.

📌 Gen - AI Architect (Chennai)
🏢 Johnson Electric
📍 Chennai