Vice President of Artificial Intelligence

You will run AI at our client. Our platform turns raw, multi-source content into a temporal knowledge graph of concepts, and we are strong on the symbolic side — reasoning, ontologies, extraction logic. What we have not yet hired is the leader who owns the learned half of the system and the function that builds it: the models that let the graph construct, link, and evolve itself, the inference that serves them on our own infrastructure, and the team that will scale all of it.

This is a player-coach role, and we mean both words. You set the AI roadmap, hire and grow the team, and are accountable for what the AI layer ships. You also still write the hardest code yourself — the GNNs, the inference path, the eval harness. We are a flat, ~10-person company; a VP here is the most technical person in the room, not the least.

The stakes are concrete. Three of our six core hypotheses — Autonomous Graph Construction, Concept Extraction Superiority, and Graph of Thought — stay unproven until someone can train, evaluate, ship, and lead the work behind them. You own that. If the learned layer is weak, or can only run on someone else’s API, the differentiation is theoretical. You make it real, make it ours, and build the team that keeps it ahead.

The Mandate

• Own AI direction: the technical direction for the learned layer — graph ML, inference, and the production pipeline — sequenced against the patent claims and the six hypotheses.
• Build the team: hire the AI engineers under you, level them, and grow the people already here (including mentoring our junior AI engineer on the pipeline). You build the bench.
• Own the shared seam: the pipeline is where symbolic and learned work meet. You hold that shared surface and partner with our senior symbolic / reasoning engineer rather than fence off territory.
• Set the engineering bar: set the bar for evaluation, model independence, and production readiness, and make the whole AI team ship to it.

What You Will Own — And Still Build

Pillar 1 — Graph Learning & Representation

• Build GNN-based link prediction to predict missing, typed relationships between concept nodes over the live graph, replacing today’s heuristic and similarity-only linking.
• Design concept, node and edge embeddings that feed concept resolution and relationship discovery, turning identity and linkage into learned, confidence-weighted judgments.
• Use the Graphiti temporal graph to detect drift and the shift-vs-split signal from how a concept’s neighbourhood changes over time, not from rules alone.

Pillar 2 — Inference & Production Systems

• Stand up provider-independent inference: open-weight models on Groq / LPU behind a stable, swappable interface, with an in-tenant GPU path for sensitive tenants.
• Harden the extraction pipeline behind the gRPC / Protobuf CognitionOS API — latency budgets, throughput, back-pressure, observability.
• Build the eval / regression harness that gates merges across the whole AI team — the safety net everyone builds on.

How We Work

Flat team. We don’t over-define reporting lines on a ~10-person team; seniority and ownership sort that out in practice. You come in as the senior AI leader and earn the org around you.

Location-flexible. London, Dubai, or remote works for the right person, with direct access to the founder from day one. Our design partners in legal, consulting and accounting are in the GCC, so some travel to the region matters as the AI layer meets product, but relocation is not a condition of the role for now.

Why This Role

You run a half of the thesis nobody else can. The team is deep on symbolic reasoning. The learned graph — GNNs, embeddings, temporal ML — is open, load-bearing for the patents and the product, and yours to own and staff.

Lead without losing your hands. Most VP roles take you off the keyboard. This one is the opposite — the value is in a leader who still ships the hardest model and inference code while building the team that scales it.

Build the function from the front. You’re not inheriting a large org with its habits set. You set the bar, hire to it, and decide what the AI team becomes.

9 patents, 119 claims. Novel, defensible architecture with a live system already serving design partners — not a wrapper around someone else’s API. Your leadership and your hands both have a real home here.