Founding Agentic Reasoning Engineer

Founding Agentic Reasoning Engineer
Applied AI | Build the Brain
Our client is building a system where AI doesn’t just generate answers. It reasons, validates, and defends them using structured evidence.
They’re hiring a Founding Agentic Reasoning Engineer to design and own the first real reasoning layer powering that system.
This is where complex questions get broken down, evidence gets challenged, and answers stand up to scrutiny.
What you’ll do
Architect multi-step AI reasoning workflows. From question → decomposition → retrieval → answer
Build tool-using agents that decide what to fetch, what to trust, and how to resolve conflicting data
Design systems that handle uncertainty, contradictions, and missing context properly
Combine internal structured data with external sources, and decide when each matters
Create reasoning traces that make system decisions transparent and debuggable
Build and iterate on benchmarks + evaluation loops that prove the system is improving
Turn one-off demos into repeatable, reliable workflows
Work closely with data teams to shape how evidence is structured for better reasoning
What they want
You can take a messy, ambiguous question and break it into clear, solvable steps
Strong understanding of agentic / multi-step AI systems (retrieval, routing, synthesis)
You know where AI breaks and how to fix it (not just patch it)
You think in terms of systems, not prompts
Comfortable working with uncertainty, conflicting data, and incomplete information
You can build and debug what you design
Nice to have
Experience with RAG, knowledge graphs, or structured data systems
Exposure to evaluation frameworks, benchmarking, or human-in-the-loop setups
Familiarity with modern reasoning techniques (RL, distillation, etc.)
Background in high-stakes domains where accuracy and provenance matter
Why this role
Own the reasoning layer of an AI system from day one
Work on problems where correctness actually matters
Shape how AI systems think, not just what they say
What it’s not
Not a pure research role
Not prompt engineering
Not a generic chatbot build
Show more Show less