Your Impact at LILA

Lila Sciences is building the infrastructure for autonomous science: AI systems that design experiments, run them in physical labs, and close the loop across life sciences, chemistry, and materials science. Our platform enables this to operate reliably at scale for external customers who require security, auditability, and operational rigor.

We're hiring a Staff/Principal Engineer, Search & Retrieval (Enterprise AI) to own how scientists and agents ask questions of scientific knowledge at Lila: published literature and patents, customer-supplied documents, lab protocols, and the structured experimental record.

Your work will involve building RAG over unstructured text and a scientific knowledge graph that links literature to the structured experimental record.

Enterprise customers require one retrieval platform that any scientific workload can build on, with tenancy enforced in the index and the graph rather than the application layer. You will own that stack end to end, from ingestion to retrieval to cited answers a scientist can trace back to the evidence.

This is a deeply technical senior individual contributor role. We move fast, ship often, and look for engineers who do well with ambiguity and enjoy shaping the problem as much as solving it.

What You'll Be Building

Technical leadership across platform engineering, infrastructure, AI/ML, and scientific teams. Broad architectural influence, high autonomy, direct ownership of critical platform systems.
Define the shared retrieval foundation. Your architecture decisions become the default for scientific workloads across the company, well beyond the first one you ship.
Knowledge graph and graph retrieval. Build one scientific knowledge graph from literature and the structured experimental record. Own graph retrieval and its fusion with vector and lexical search into a single cited answer.
Agentic search. Build the autonomous research loop, from query planning through retrieval to traceable reasoning.
Retrieval quality. Own the retrieval pipeline end to end and the evaluation harness.
Ingestion and extraction. Own the pipeline that turns the many types of scientific data (PDFs, patents, lab protocols, and more) into semantically searchable evidence.

What You'll Need to Succeed

8+ years of software development experience, including production ownership of a search, retrieval, RAG, or knowledge graph system that real users depended on.
Production experience building semantic search: embedding models, vector indexes (HNSW/IVF), hybrid lexical and vector retrieval, and reranking. Graph database experience is a strong plus.
Track record in high-growth environments where speed-to-value mattered as much as long-term architecture.
You can describe how you measured retrieval quality (recall, groundedness, citation accuracy) and how you caught regressions.
Hands-on coding in Python and modern backend frameworks (FastAPI, async Python).
Modern cloud-native architectures (Kubernetes, Terraform, AWS).
Strong cross-functional communication and the ability to influence architecture across teams.
BS, MS, or Ph.D. in Computer Science or a related field of study, or equivalent experience.

Bonus Points For

Shipped RAG or semantic search in a regulated or enterprise context, through security and compliance review under regimes such as SOC 2, HIPAA, GxP (including ALCOA+ data integrity), or FedRAMP.
Permission-aware retrieval over access-controlled data.
Document understanding at scale: OCR, layout-aware parsing, table and figure extraction from scientific literature.
Agentic retrieval and GraphRAG patterns, and how to evaluate them.
Platforms serving scientific or laboratory data such as LIMS, ELN, bioinformatics, or materials informatics, with an understanding of provenance and traceability requirements.
Domain-specific corpora and scientific ontologies (UMLS, MeSH, ChEBI, GO, UniProt, Materials Project) where specialized vocabulary limits general-purpose embeddings and entity linking.
Knowledge graph construction and graph retrieval: property graph or RDF stores.