Lead Data Engineer (Hands-On)

Lead Data Engineer (Hands-On)

Cephas Consultancy Services Private Limited Cary, North Carolina, United States

About this position

Client:

Employment: Full-Time

(On-site / Hybrid)

Salary: $ - $ per annum plus benefits
Eligibility: US Citizens only

ABOUT THE ENGAGEMENT

A centralized, AI-first enterprise Data Hub for a global insurance and financial services client on Azure Databricks. The platform ingests 150+ inbound data feeds, distributes to 35+ downstream systems, and is organized as a medallion architecture (Bronze / Silver / Gold). AI is embedded in ingestion, canonical mapping, data quality, reconciliation, and business user access from day one.

This is a senior hands-on leadership role. The candidate will own the end-to-end technical design of the data and AI layers, build reference implementations for the engineering team, and ship production-grade Python, Scala, and PySpark code every week. Candidates who have not written or reviewed production code in the past year are not a fit.

WHAT THE ROLE OWNS

• Data platform architecture and engineering: lakehouse architecture (Bronze / Silver / Gold contracts, ADLS Gen2 zone layout, Delta Lake table design, partitioning, schema evolution, retention).
• Metadata-driven, parameterized ingestion frameworks for batch files, database extracts, CDC feeds and streaming (Azure Event Hubs / Kafka,
  Spark Structured Streaming).
• Canonical PySpark and Scala Spark jobs, coding and testing standards, PR reviews, production incident debugging, Spark cluster tuning and cost guardrails.
• CI/CD for Databricks and ADF in Azure DevOps using Databricks Asset
  Bundles and Terraform; observability with Azure Monitor and Log
  Analytics.
• AI-augmented ingestion and canonical mapping: auto-generated bridge documents, DML, canonical table definitions; AI-assisted source-to-canonical mapping with human review gate.
• AI-driven data quality, anomaly detection (data drift, schema drift,
  volume shifts, reconciliation breaks), automated reconciliation, and
  synthetic privacy-preserving test data.
• Semantic layer and knowledge graph, plus a GPT-powered
  conversational interface (text-to-SQL / semantic-layer retrieval) with
  row- and column-level security.
• Governance and leadership: Unity Catalog (lineage, access control,
  PII standards), Architecture Review Boards and AI governance forums,
  mentoring engineers, documentation.

MUST-HAVE SKILLS & EXPERIENCE

• Expert-level Python, Scala and PySpark: production-ready, modular,
  well-tested solutions; Spark workload troubleshooting; optimizing
  large-scale batch and streaming pipelines using Delta Lake.
• Strong SQL and data modelling (dimensional and normalised), schema
  design, data contracts.
• Databricks expertise: Delta Lake, Unity Catalog, Jobs & Workflows,
  cluster and pool management, performance tuning, Model Serving.
• Azure data stack: ADLS Gen2 (zone design, ACLs, lifecycle), Azure
  Data Factory (parameterized / metadata-driven frameworks), Azure Event
  Hubs.
• 3+ years designing and shipping LLM-based systems in production: RAG
  pipelines, agentic / tool-calling workflows, chunking and embedding
  strategy, vector and hybrid retrieval, prompt engineering.
• Evaluation discipline: golden datasets, regression suites, accuracy
  and hallucination tracking, human-in-the-loop feedback.
• Hands-on with LangChain, LlamaIndex or LangGraph, plus at least one
  provider stack (Azure OpenAI, OpenAI, or Databricks Model Serving).
• Metadata-driven frameworks: schema inference, data profiling,
  lineage, catalogs.
• 12-18 years of total experience in data engineering / data platform delivery.
• Proven enterprise-scale delivery of a medallion / lakehouse architecture.
• Azure security and governance: Entra ID, managed identities, RBAC,
  POSIX ACLs, Key Vault, private endpoints, PII handling.
• CI/CD and IaC: Azure DevOps, Terraform, Databricks Asset Bundles,
  automated testing of data pipelines.
• Clear technical writing and ability to present and defend designs to
  engineers and non-technical stakeholders.

STRONGLY PREFERRED

• Knowledge graphs and ontologies (RDF/SPARQL, Neo4j, graph modelling
  over a lakehouse).
• Text-to-SQL or semantic-layer-backed natural-language query systems
  at enterprise scale.
• ML-based anomaly detection on time-series or transactional financial data.
• Financial services or insurance domain (finance close, GL,
  subledger, reconciliation, actuarial data).
• LLMOps / MLOps: model and prompt versioning, cost governance, observability.
• Databricks Data Engineer Professional, Azure DP-203 / DP-700, or
  AZ-305 certification.
• dbt, Great Expectations or similar; Workday, Prism or Accounting
  Center exposure.