Senior ML Scientist

About Katalyze AI

Katalyze AI is a fast-growing AI-driven biotech platform company on a mission to make life-saving drugs accessible and affordable for everyone. Our AI Agents help pharmaceutical and biotech companies increase production efficiency, reduce costs, and minimize waste. We're a team of humble, fast-moving, and curious craftspeople working at the intersection of science and AI.

About the Role

Katalyze AI is looking for a Senior ML Scientist to design, build, and deploy the core intelligence layer of our platform. You'll work across two interconnected domains: RAG and knowledge retrieval (making our platform reason accurately over large scientific document corpora) and agentic AI systems (building the multi-step reasoning pipelines that autonomously complete complex tasks for our biopharma and manufacturing customers).

This is a research-forward role with direct impact on production. You'll go from reading a paper to shipping a production system; sometimes in the same week.

• Design and build advanced RAG pipelines for scientific knowledge retrieval: chunking strategies, embedding model selection, hybrid search, re-ranking, and rigorous retrieval evaluation

• Develop and maintain Knowledge Graph architectures (Neo4j, ontologies, semantic structures) that capture domain relationships and give agents deep understanding of biopharma and manufacturing workflows

• Architect agentic workflows using LangChain/LangGraph or custom orchestration: designing autonomous, multi-step reasoning pipelines for complex enterprise tasks

• Build the "skills layer" that allows agents to execute domain-specific tasks reliably, with proper validation, auditability, and error handling for high-stakes regulated environments

• Advance entity extraction and knowledge representation: building systems that turn unstructured scientific documents into structured, queryable domain knowledge

• Design and run rigorous evaluation frameworks to benchmark agent reliability, RAG accuracy, and model consistency — define what "good enough to ship" looks like

• Stay current with ML research (NeurIPS, ICML, ICLR, ACL) and identify applicable advances; translate them from paper to production

• Collaborate with the Data Science, Engineering, and Product teams to integrate ML components into customer-facing features

What We're Looking For

• 4+ years of applied ML research or engineering experience, with production deployments under your belt

• Deep RAG expertise: chunking, embedding models, vector databases (Pinecone, Weaviate, pgvector), hybrid retrieval, context window optimization, and evaluation methodology

• Hands-on experience with Knowledge Graph construction (Neo4j, RDF/OWL, property graphs) and graph-based reasoning

• Proficiency with agent frameworks: LangChain, LangGraph, AutoGPT, CrewAI, or custom orchestration — and real opinions on their limitations

• Strong LLM engineering skills: prompt engineering, structured output validation, token cost optimization, API integration, handling non-determinism in production

• Production-grade Python engineering: clean code, testing (including golden-dataset eval pipelines), CI/CD, version control

• Experience with PyTorch or JAX; familiarity with Hugging Face ecosystem

• PhD or Master's in Machine Learning, NLP, Computer Science, or related field preferred

• Domain knowledge in life sciences, biopharma, chemistry, or industrial processes is a significant advantage

Tech Stack:

• Agent Frameworks: LangChain, LangGraph, custom orchestration

• LLM Providers: Anthropic Claude, OpenAI, AWS Bedrock

• Knowledge Systems: Neo4j, custom ontologies, semantic search, pgvector / Pinecone

• ML & Research: Python, PyTorch, Hugging Face, scikit-learn, pandas

• Infrastructure: AWS, Docker