Lead - AI Engineering (ML, Evaluation & Databricks)

Role Details: Leads the engineering of production-grade AI and ML solutions on Azure Databricks, with accountability for model development, applied statistics, evaluation, lifecycle management and operational quality.
Job Location : Bangalore/Gurugram/Mumbai
OVERALL JOB DESCRIPTION :
AI/ML Solution Engineering
Lead the end-to-end engineering of AI and machine-learning solutions, from problem formulation and technical design through data preparation, model development, deployment, monitoring and ongoing improvement.
Design modular and reusable components for feature engineering, training, evaluation, inference, model access, feedback, monitoring and governance rather than building isolated notebook-based solutions.
Partner with AI Application Engineering teams to integrate predictive models, retrieval components, evaluation services and GenAI capabilities into secure enterprise applications.
Determine when predictive ML, statistical methods, search, rules, GenAI or a combined engineering approach is most appropriate.
Machine Learning & Applied Statistics
Engineer production ML solutions across classification, regression, ranking, forecasting, clustering and anomaly detection use cases.
Apply statistical methods to define targets, sampling strategies, baselines, validation design, metrics, calibration, thresholds and release criteria.
Identify leakage, bias, class imbalance, confounding, temporal effects and uncertainty and conduct systematic error analysis across relevant business segments.
Balance model performance with explainability, stability, latency, operational complexity, maintainability and cost.
AI Evaluation Engineering
Build reusable evaluation frameworks for predictive models, retrieval systems, LLM applications and bounded agentic workflows.
Create representative labelled and regression datasets covering normal, difficult, edge and adversarial cases and establish measurable production-readiness thresholds.
Combine deterministic measures, statistical analysis, human review, custom scorers and appropriately calibrated model-based judges.
Separate and diagnose failures across data preparation, retrieval, generation, model inference, tool selection and workflow execution.
Azure Databricks, MLflow & Unity Catalog
Build governed AI workloads using Python, SQL, PySpark, Delta Lake, Databricks Runtime, Workflows, Model Serving, Vector Search, MLflow and Unity Catalog.
Use MLflow for experiment and application tracking, evaluation, tracing, packaging, model registration, versioning, promotion, deployment, rollback and monitoring.
Govern data, features, models, prompts, functions, evaluation datasets and AI artefacts through Unity Catalog access control, lineage and auditability.
Integrate Databricks-hosted models and services with Azure-hosted applications through secure APIs, identity and approved network patterns.
Production AI Engineering & Operations
Convert exploratory notebooks into modular, versioned, tested and deployable Python packages, services and Databricks workflows.
Implement CI/CD, automated data and model testing, configuration management, environment promotion, release controls and reproducible builds.
Deploy batch, streaming and real-time inference and optimise compute, latency, throughput, reliability and cost.
Monitor data quality, feature and prediction drift, model and application quality, serving health and business outcomes; define retraining, rollback and retirement triggers.
Technical Leadership, Risk & Delivery
Lead and mentor AI/ML engineers and applied data scientists while remaining hands-on in architecture, coding, model reviews, evaluation design and production issue resolution.
Define AI engineering standards, reusable patterns, acceptance criteria and production-readiness gates for model and evaluation components.
Partner with AI Product Managers, Application Engineering, Data Engineering, Enterprise Architecture, Information Security, Risk, Model Risk, Compliance and Audit.
Challenge vendors and implementation partners on model claims, benchmarks, evaluation methods, platform design and production quality.
EDUCATION :
Bachelor’s or Master’s degree in Computer Science, Engineering, Statistics, Mathematics, Data Science, Econometrics, Operations Research or a related quantitative discipline.
Strong production engineering, machine-learning and applied statistical competence is more important than research credentials.
Preferred: Recognised Databricks, MLflow, Azure, cloud architecture or machine-learning engineering certification.
WORK EXPERIENCE :
9-12 years of overall technology experience.
5+ years building AI/ML solutions, including strong hands-on Python development, applied modelling and production-system ownership.
4+ years in software, data or ML engineering, including conversion of exploratory work into tested and deployable production components.
3+ years with Databricks, Apache Spark or a comparable lakehouse and distributed processing environment.
2+ years in Generative AI, NLP, RAG, document intelligence or evaluation of LLM and agentic applications.
3+ years deploying and operating ML systems, inference pipelines or AI services on Azure or another major cloud platform.
2+ years leading engineers, data scientists or complex technical workstreams while remaining actively involved in architecture, coding and reviews.
Demonstrated ownership of at least two AI/ML solutions from problem definition and model design through deployment, monitoring and ongoing support.
TECHNICAL / FUNCTIONAL SKILLS :
AI/ML Engineering
Advanced Python, modular and object-oriented design, typing, packaging, dependency management, configuration-driven development and performance optimisation.
Production model pipelines covering data preparation, feature engineering, training, validation, packaging, deployment and inference.
API and service integration, batch/streaming interfaces, error handling, retries, idempotency and operational resilience.
Git, peer review, secure coding, documentation and unit/integration/data-quality/model-regression testing.
Machine Learning
Classification, regression, ranking, recommendation, clustering, forecasting and anomaly detection.
Feature engineering and selection, model comparison, hyperparameter optimisation, ensemble methods and error analysis.
NLP, embeddings and transformer fundamentals and practical judgement on pre-trained models, fine-tuning and custom models.
Explainability, robustness, stability, fairness, model limitations and human-review requirements.
Applied Statistics & Experimentation
Probability distributions, sampling, descriptive and inferential statistics, hypothesis testing, confidence intervals and regression.
Experimental design, A/B testing, statistical power, sample-size considerations and practical versus statistical significance.
Calibration, threshold optimisation, class imbalance, leakage, confounding, temporal validation, uncertainty and causal pitfalls.
Ability to translate statistical findings into engineering decisions and measurable business outcomes.
Databricks Engineering
Python, SQL, PySpark, Delta Lake, lakehouse patterns and distributed data processing.
Databricks Runtime for ML, notebooks/repos, Jobs, Workflows and deployment bundles or equivalent release practices.
Feature computation and management, batch and streaming pipelines and workload performance/cost optimisation.
Model Serving, scalable batch and real-time inference, Vector Search and Databricks APIs/SDKs.
MLflow, Unity Catalog & Lifecycle
Experiment and application tracking, parameters, metrics, artefacts, packaging and reproducibility.
Models and AI artefacts in Unity Catalog, versions and aliases, promotion, rollback, access control, lineage and auditability.
MLflow tracing, evaluation datasets, scorers, feedback, prompt/application versions and production monitoring.
End-to-end traceability across source data, features, code, models, prompts, tools, configurations and evaluation results.
GenAI, RAG & Agent Evaluation
Working knowledge of LLM behaviour, model APIs, structured outputs, context management, RAG and tool-calling patterns.
Retrieval evaluation using precision/recall at K, ranking relevance, labelled query-document datasets and failure segmentation.
Generation evaluation covering groundedness, correctness, completeness, consistency, robustness, safety and task completion.
Deterministic checks, custom scorers, human evaluation, regression datasets and calibrated LLM-as-judge approaches.
Agent evaluation covering tool-selection accuracy, workflow trajectory, execution outcome, approval controls and failure recovery.
Azure, Deployment & Operations
Azure identity and security concepts including Entra ID, managed identities, Key Vault, private networking and API access.
Integration with Azure-hosted applications, A

Work arrangement
No