Machine Learning Engineer (Platform)

Location: Hybrid in the San Francisco Bay Area or New York City, NY

Company

hirify.global builds AI-native operating systems for large, regulated institutions across government, insurance, health, financial services, and construction.

What you will do

• Build and own shared ML capabilities for the Atlas platform, including document extraction, foundation models, model routing, and unified evaluation.
• Develop composite AI systems combining vision models, VLM reasoning, LLMs, agents, and rule engines.
• Turn project-pod needs into reusable platform capabilities without over-abstracting before patterns are proven.
• Engineer continuous improvement loops that capture production corrections, identify component failures, and support retraining and safe redeployment.
• Optimize production systems across accuracy, latency, cost, reliability, and changing deployment environments.
• Serve internal project pods and domain experts as customers while transferring platform learnings across products.

Requirements

• Strong understanding of machine learning fundamentals, including loss functions, generalization, distribution shift, and evaluation.
• Hands-on experience with modern AI systems, including LLM prompting, fine-tuning, tool use, and reasoning.
• Ability to choose and compose fine-tuned models, vision-language models, and rule engines for specific tasks.
• Experience building production ML or agentic systems with measurable quality, latency, cost, and reliability requirements.
• Ability to identify reusable capabilities across multiple teams and own them end-to-end in production.

Culture & Benefits

• Work on applied AI systems deployed to production for large regulated institutions.
• Build new platform capabilities at the research frontier with direct customer impact.
• Collaborate with an engineering team formed from backgrounds including Palantir, Google, Meta, and Nvidia.
• Work on systems that compound institutional intelligence through verified corrections and feedback loops.