AI Data Architect

3Pillar

11d•Hybrid

About The Position

We are looking for an AI Data Architect to design, build, govern, and evolve the single source of truth that powers every AI initiative in our organization. This platform will serve as the foundational nervous system for conversational AI assistants, dashboard intelligence, autonomous AI agents, RAG-powered applications, predictive ML models, and any AI product we build today or in the future. The resource will architect the system, drive implementation, own the data contracts that agents and AI applications depend on, enforce security and access governance for both human and agent consumers, and continuously monitor and improve the accuracy and reliability of AI outputs that flow from this platform.

Requirements

Python
SQL
Snowflake/Databricks
AWS (S3, Glue, EKS, Bedrock, Kinesis, Redshift)
Docker
Kubernetes
Terraform
GitHub Actions
LangChain
LlamaIndex
LLM APIs (OpenAI, AWS Bedrock, Claude, HuggingFace)
(Pinecone, FAISS, ChromaDB, OpenSearch)
knowledge graphs (Neo4j)
MLflow
FastAPI
CI/CD pipelines
observability tooling (CloudWatch, Grafana, or equivalent)
data lineage and metadata management platforms
15+ years of hands-on data engineering and architecture experience, alongside building production AI/ML and LLM-era data infrastructure.
Strong Experience with either Databricks or Snowflake; experience with both is desirable.
Strong data architecture patterns & principles, ability to design secure & scalable data lakes, data warehouse, data hubs, and other event-driven architectures
Expertise in designing and writing ETL processes in Python / Java / Scala
Own the full data stack: real-time streaming (Kafka, Spark Structured Streaming), batch processing (Databricks, PySpark, Delta Lake), cloud storage and compute (AWS, Azure), and data quality /metadata management.
Drive modernisation of legacy pipelines (on-prem ETL, batch DWH) to cloud-native, AI-ready architectures with measurable improvements in cost, latency, and delivery velocity.
Proven experience designing enterprise-scale AI data platforms that serve multiple AI consumers â€”not just one application or pipeline.
Hands-on experience with vector stores, semantic models, knowledge graphs, and retrieval infrastructure in production environments.
Working knowledge of LLMOps: model serving pipelines, MLflow, CI/CD for AI, automated evaluation, and production monitoring.

Nice To Haves

experience with both Databricks and Snowflake is desirable

Responsibilities

Architect and own the enterprise AI data platform â€” the unified, governed layer that ingests, transforms, stores, and serves all data consumed by AI systems across the organisation.
Design multi-domain data models (lakehouse, data mesh, event-driven) that are structured from day one to serve AI workloads: clean lineage, versioned schemas, well-documented contracts, and low-latency serving APIs.
Define tools & technologies to develop automated data pipelines, write ETL processes, develop dashboard & report and create insights
Design the retrieval infrastructure that powers RAG-based AI applications: embedding pipelines, vector stores (Pinecone, FAISS, ChromaDB, OpenSearch), chunking strategies, and hybrid retrieval layers combining semantic search with structured queries.
Own the observability stack for AI agent behaviour: instrument agents to capture inputs, retrieved context, tool calls, reasoning traces, and outputs â€” creating a complete audit trail of every agentic action driven by platform data.
Design and operate evaluation frameworks that continuously measure AI output quality: factual accuracy, context faithfulness, retrieval relevance, hallucination rates, and task completion successâ€” across all AI consumers of the platform.
Define and maintain the reference architecture for the AI data platform â€” documenting design patterns, data contracts, integration standards, and decision records (ADRs) that all engineering teams follow.
Establish data engineering standards: pipeline testing frameworks, code review practices, CI/CD automation, infrastructure-as-code (Terraform), reusable component libraries, and observability instrumentation.