AI Architect

We are seeking a seasoned AI Architect to define and own the intelligence layer of our AI- native automotive retail platform. This is a deeply applied, Python-first role that sits at the intersection of product strategy, data science, and system architecture. You will design the end-to-end ML system — from Kafka-based feature ingestion and Python model training through Redis/Aerospike serving and LLM gateway integration — and architect the hexagonal adapter patterns that connect AI capabilities to Java backend services without coupling domain logic to infrastructure.

Responsibilities:

• Design and own intelligent scoring models in Python: customer propensity models, vehicle equity scoring, service retention prediction, and opportunity ranking — consuming MongoDB and Kafka event data.
• Architect the AI-native ML pipeline: Kafka feature stream ingestion, Python feature engineering, model training (scikit-learn, XGBoost, PyTorch), Elasticsearch-backed feature stores, and Redis/Aerospike hot-path serving layer.
• Define the hexagonal adapter pattern for AI integration: port interfaces in Java domain services that abstract LLM gateway calls, Redis/Aerospike score reads, and ML pipeline writes — keeping domain logic AI-infrastructure-agnostic.
• Lead experimentation and rapid prototyping: run continuous Python-based POCs in close collaboration with Product and Engineering to validate ideas quickly and de-risk investments.
• Build and evolve recommendation systems, propensity models, ranking algorithms, and next-best-action engines that integrate seamlessly via well-defined API contracts with Java microservices.
• Evaluate emerging AI/ML techniques and GenAI capabilities (RAG, fine-tuning, agentic patterns) —determine practical applicability, integration cost, and ROI for the platform.
• Establish model lifecycle best practices: evaluation, monitoring, retraining triggers, bias detection, and explainability that build dealer trust.
• Shape long-term AI architecture: feature stores, model serving patterns (Redis/Aerospike vs. REST API), Cosmos DB event log integration, and LLM gateway usage policies.
• Communicate complex ML concepts to senior stakeholders; mentor engineers on applied ML design and AI-native engineering patterns.

Requirements:

• Deep Python expertise: scikit-learn, XGBoost/LightGBM, PyTorch, Pandas, and production ML tooling.
• Proven experience designing recommendation engines, propensity scoring systems, ranking algorithms, or personalisation platforms at production scale.
• Strong experience building ML feature pipelines that consume Kafka event streams and store features in Elasticsearch, Redis, Aerospike, or MongoDB.
• Ability to design hexagonal adapter interfaces between Java backend services and Python ML serving layers.
• Excellent Java and object-oriented development skills.
• Experience in building and scaling microservices.
• Solid understanding of at least one NoSQL database.
• Strong problem-solving skills, technical troubleshooting and diagnosing.
• Expected to be a role model for young engineers, have a strong sense of code quality, and enforce code quality within the team.
• Experience with Redis or Aerospike for sub-10ms model score serving.
• Comfortable communicating complex ML ideas to non-technical stakeholders; pragmatic and outcome-driven.

Good to Have:

• Experience with LLM/GenAI integration: RAG architectures, prompt engineering, LLM gateway patterns, and agentic workflow design.
• Familiarity with Cosmos DB for globally distributed ML event logging and experiment metadata.
• Background in martech, customer intelligence, or real-time personalisation platforms.
• Experience with distributed ML training (Spark MLlib, Ray, or equivalent).