AI/ML Engineer - NLP Scientist

Dawar Consulting•South San Francisco, CA

8d•Onsite

About The Position

Our client, a world leader in biotechnology and life sciences, is looking for a Senior AI/ML Engineer - NLP Scientist. This role involves building an evidence-grounded AI capability that verifies generated claims against approved scientific, clinical, regulatory, and reference materials before human review. The system will retrieve relevant evidence, decompose claims into verifiable assertions, evaluate evidence support, and provide traceable decisions with citations. The system must recognize unsupported or contradicted claims and abstain rather than guess.

Requirements

Strong Python production engineering.
NLP / LLM / Generative AI development.
RAG, hybrid search, vector search, and lexical retrieval.
Natural Language Inference (NLI), entailment, contradiction detection.
Claim decomposition and evidence attribution.
LLM/model APIs and production evaluation frameworks.
AI/ML evaluation, benchmarking, and error analysis.
Human-in-the-loop AI, confidence scoring and abstention.
Experience with scientific, technical, regulatory, legal, or other high-stakes content.
Experience creating expert-labeled datasets and annotation guidelines.
Strong understanding of traceability, citations, and reproducible AI decisions.

Nice To Haves

Knowledge graphs and relationships between claims, evidence, references, products, and indications.
Deterministic rules + ML/LLM decision systems.
Pharmaceutical, biotech, healthcare, regulatory, legal, financial compliance, or scientific publishing experience.
Familiarity with clinical studies, statistics, scientific literature, and citation practices.
Experience with LangChain, LlamaIndex, Hugging Face, PyTorch, or similar NLP/ LLM frameworks.

Responsibilities

Build production-grade Python/NLP pipelines for claim verification and evidence attribution.
Develop hybrid retrieval using lexical and vector search to identify relevant evidence.
Implement claim decomposition, natural language inference (NLI), entailment, and contradiction detection.
Evaluate whether generated claims are genuinely supported by cited evidence.
Design confidence thresholds, abstention logic, escalation rules, and human-in-the-loop workflows.
Build evaluation datasets with expert annotation guidelines and measure inter-annotator agreement.
Track false approvals, false rejections, abstentions, and other error categories.
Develop traceable systems that allow decisions to be reconstructed based on model version, evidence, citations, and reviewer actions.
Work with Medical, Legal, Regulatory, and scientific stakeholders to translate review requirements into technical solutions.