Principal AI Architect

Draup is a Series A-funded agentic AI company building the intelligence layer for how global enterprises make workforce and go-to-market decisions. We work with 250+ enterprise clients — including 5 of the Fortune 10 — processing 1B+ job descriptions, 850M+ professional profiles, and signals from 100+ labor databases.

We are now building our Silicon Valley engineering team — a small, senior group focused on next-generation AI research and product.

Location: San Francisco, SoMa — minimum 4 days per week in-office.

What you'll do

• Define and own the architectural vision for next-generation AI systems: novel agent architectures, reasoning systems, and proprietary model development.

• Prototype and evaluate breakthrough AI capabilities — predictive modeling, autonomous signal synthesis, multi-modal intelligence — that create defensible IP.

• Set the technical foundations: data architecture, model serving, agent orchestration, and inference infrastructure for a multi-year product horizon.

• Evaluate and adopt emerging AI paradigms (multi-agent reasoning, RLHF, retrieval-augmented fine-tuning) before they are mainstream.

• Establish engineering standards, architecture review processes, and IP documentation practices for the team.

What we require

• BS/MS/PhD in Computer Science, AI/ML, or related field. PhD or equivalent research depth preferred.

• 5+ years in AI/ML engineering with at least 2 years in a principal engineer or lead architect role.

• Demonstrated history of building original AI systems that became products — not implementations of existing patterns.

• Deep expertise across: LLM training and fine-tuning, agentic system design, knowledge graph construction, large-scale data modeling.

• Comfort operating in greenfield conditions: high ambiguity, high ownership.

• Patents, published research, or shipped products with clear proprietary differentiation are strong signals.

• No visa sponsorship. Must be authorized to work in the US without current or future employer sponsorship.