Founding AI Engineer - CompoundX

About CompoundX

The plastic compounding market is worth $78B - and it still runs on PDFs, email threads, and expert intuition locked in people's heads. CompoundX is building the AI OS for material formulation: encoding that expert knowledge, automating brief translation, and routing production orders to the right specialists.

We are pre-PMF, iterating fast. Founded by 2 serial entrepreneurs in June 2026, €1M pre-seed raised, 3 clients among Europe's largest industrial brands, and a partnership with a top-10 global plastic compounder. Backed by OSS Ventures - France's leading industrial AI builder. 24+ companies created, 3,000+ plants, 500+ employees across the portfolio.

Paris today. The market is global. The ambition is to become the autonomous material R&D layer for industrial brands worldwide.

---

---

The Role

Ali, CTO & Co-founder, is simultaneously architecting the core system, shipping product, and delivering to clients. That rhythm is not sustainable alone. We need a second engineering pillar - someone who owns a full slice of the build and holds the technical bar without being managed at every step.

You are not a resource. You are the co-builder who makes it possible to move fast without breaking what matters.

---

What You'll Do

• Design and ship agentic AI systems - multi-step workflows that reason, act, and recover from failure in real production environments

• Build the knowledge layer - ingest formulation specs, technical datasheets, and material constraints into structured, reliable AI memory

• Own evaluations - design feedback loops and benchmarks that tell us whether the AI is actually getting better with each iteration

• Encode expert knowledge - sit with a material engineer, understand how they think, and translate that into a data model. This is the hardest and most valuable part of the job.

• Deliver end-to-end - from feature to production to client deployment. You sit with the client, see what works, bring it back to the next sprint.

• Shape the product - challenge specs, flag weak ideas, propose better ones. You are early enough that what you build defines what CompoundX becomes.

---

What We're Looking For

Must-haves

• 5–8 years full-stack experience, strong backend bias - we care about what you've shipped, not the year count

• LLM systems in production - not prototypes. Agentic workflows: orchestration, tool use, state management, failure handling

• You think in evals - you know how to measure whether an AI system is actually improving

• Engineering mindset - you think in systems, constraints, and failure modes. You are not afraid to destroy in order to rebuild better. Coding is your tool; engineering is your mindset.

• Domain curiosity - you can sit with an industrial expert, ask smart questions, and connect domain logic to technical architecture. You don't need to be a chemist. You need to want to understand one.

• Startup DNA - you've made architectural decisions with limited guidance and lived with the consequences

• User proximity - you've talked to users, changed what you built because of those conversations, and are comfortable in a room with an industrial engineer

• Location - Paris-based or willing to relocate. The team works on-site.

Nice-to-haves

• Experience in industrial software, materials science tooling, or PLM/ERP adjacency

• Exposure to ontology design, knowledge graphs, or structured domain modeling

• Prior experience as first or early engineering hire

---

Tech Stack

• Backend - Python

• LLM inference - OpenAI & Anthropic APIs (zero-training, zero-retention policy)

• Infrastructure - Google Cloud Platform (GCP), Europe region

• Architecture - Multi-tenant, client data isolation, RBAC, AES-256 encryption at rest, TLS 1.3 in transit

• CI/CD - Automated pipeline with security testing (Dependabot, Snyk)

• Frontend - Next.js 16 (React 19, TypeScript), Tailwind CSS, shadcn/ui, tRPC + TanStack Query, React Hook Form & Zod

---

This Is Probably Not For You If…

• You work best when requirements are fully defined before you start

• You need to own the idea to execute on it well

• You prefer deep specialization over owning the full loop from build to client

• You're looking for a stable environment with clear processes and predictable sprints

• You want to optimize one layer of the stack - here, you own several