Principal Applied Scientist

We are seeking a Principal Applied Scientist to be the highest-level AI technical authority on the team. In this role, which is above the Applied Scientist level, you set the AI research and engineering direction, own the agentic system architecture, define the eval framework standards, and operate as the primary interface to the external AI research community and internal platform AI leadership. You will combine deep research rigor with relentless production focus: every AI capability you design ships, gets measured, and improves based on evidence.

Responsibilities:

• Own the agentic system architecture: define the multi-agent coordination patterns, MCP tool taxonomy, intent-to-skill routing logic, policy enforcement design, and memory management strategy across all AI workflows on the platform.
• Lead the AI evaluation framework: design the golden dataset structure, define eval metrics (RAG retrieval quality, scoring accuracy, LLM response correctness, agentic workflow success rate), and own the CI/CD AI eval pipeline that gates every AI capability release.
• Drive the RAG architecture strategy: chunking strategies for MongoDB and Elasticsearch-indexed documents, embedding model selection and fine-tuning, retrieval reranking design, and hallucination mitigation patterns.
• Own the LLM integration architecture: prompt engineering standards, LLM gateway usage policies, context window management, streaming response patterns, and model version governance.
• Lead applied research initiatives: identify novel techniques (RLHF, Constitutional AI, multimodal, and sparse retrieval) that are practically applicable to automotive retail intelligence use cases and prototype them rigorously in Python.
• Define the hexagonal AI adapter contract standard: the formal interface specification between Python AI services and Java domain cores, ensuring AI infrastructure remains swappable without domain logic changes.
• Act as primary technical interface to platform AI leadership (LLM Gateway, ML Platform, and Agentic Engine), influencing platform roadmap based on TAMC's practical requirements.
• Mentor the Applied Scientist and ML Architect; conduct technical reviews of AI system designs, eval frameworks, and research approaches.
• Publish, present, or contribute to open source where appropriate, representing Tekion's AI engineering excellence externally.

Requirements:

• 10+ years of applied AI/ML experience with at least 3 years at the staff or principal level and a track record of shipping AI systems that measurably changed business outcomes.
• Deep expertise across: LLM fine-tuning and alignment, RAG system design (chunking, embedding, retrieval, reranking), agentic frameworks (LangGraph, CrewAI, AutoGen), and MCP tool design.
• Python mastery: PyTorch, Hugging Face Transformers, LangChain/LangGraph, scikit-learn, and production ML serving tooling.
• Strong understanding of distributed AI systems: Elasticsearch-backed vector search, Redis/Aerospike serving, Cosmos DB for AI event logging, and Kafka-based feature pipelines.
• Proven ability to design formal evaluation frameworks for AI systems, not just accuracy metrics but business-aligned quality gates.
• Experience defining and enforcing hexagonal adapter contracts between AI services and domain service layers in production systems.
• Track record of communicating complex AI system design to non-technical stakeholders, influencing product roadmaps, and driving cross-team alignment.

Good to Have:

• Publications or significant open-source contributions in NLP, information retrieval, or agentic systems.
• Experience with Constitutional AI, RLAIF, or preference optimization techniques (DPO or PPO).
• Familiarity with multimodal AI and its application to automotive retail data (image, document, and tabular data fusion).
• Prior experience in a product-focused AI role at a SaaS or platform company, not purely research.

Skills

AutoGen, CrewAI, Hugging Face, LLM, LangChain, LangGraph, MCP, PyTorch, Python, RAG, RLHF, chunking, embeddings, reranking, retrieval