Senior to Staff ML Engineer

Recruiting from Scratch is recruiting for an unnamed AI-native company building software for sensitive enterprise and national-security applications. The ML Engineer will build and own models, pipelines, agent systems, and private inference infrastructure that enable secure, reliable reasoning over complex enterprise data.

Responsibilities:

• Design and own our AI architecture, selecting, fine-tuning, and deploying the right models for each use case within controlled, private environments
• Build and maintain RAG pipelines that allow models to reason over sensitive, proprietary enterprise data without exposing it to public APIs
• Fine-tune foundation models on domain-specific data to achieve the classification accuracy and reliability our customers require
• Design and orchestrate a fleet of specialized AI agents, including agents that process and tag internal enterprise documents and agents that monitor and analyze external data sources
• Build evaluation frameworks, benchmarks, and observability tooling to ensure model reliability, accuracy, and auditability in production
• Own the private inference infrastructure, ensuring all AI runs in controlled environments that meet the security posture of our most demanding customers
• Work closely with platform and infrastructure teams to ensure AI systems are integrated securely and can scale across the product

Requirements:

• 4+ years of experience in AI/ML engineering with a strong emphasis on production systems
• Deep experience fine-tuning and evaluating large language models for domain-specific applications; you know how to get reliable, accurate results from existing models, not just benchmark performance
• Strong experience building RAG systems, retrieval architecture, chunking strategies, embedding models, and evaluation
• Experience designing and deploying multi-agent systems: orchestration, tool use, failure modes, and observability
• Hands-on experience fine-tuning (LoRA/QLoRA, full fine-tuning) and serving open-weight models locally, with a deep understanding of optimizing inference latency, throughput, and memory constraints in air-gapped or restricted infrastructure without relying on external APIs
• Strong software engineering fundamentals; you build and ship production systems, not just proofs of concept
• This role requires access to export-controlled information, as a result, applicants must be U.S. persons (citizens, lawful permanent residents, refugees, or asylees) as defined under applicable U.S. law
• Experience with information extraction, document classification, or NLP at scale is a strong plus