General
Do you work with US / non-EU clients?
Yes. We are based in the EU and have engineers across multiple time zones. Contracts can be drafted under your preferred jurisdiction with prior agreement.
What is your typical client profile?
Series A to D startups, scale-ups and mid-market companies that need a production-grade AI artifact and own the resulting IP. We do not work with adtech, surveillance, or weapons-related projects.
Do you sign NDAs?
Yes. A mutual NDA is signed before any technical discussion. We use a standard MNDA template; your own template is also acceptable.
Pricing
What is your day rate?
Our base day rate is €800 to €1,200 per engineering day, aligned with the 2026 European market for senior AI freelancers. Every engagement is billed in fixed-price phases scoped as effort × day rate.
Do you offer free trials or discounts?
No trial work. Discovery calls and proposals are always free. We do not discount published ranges — the price reflects the engineering bar and the deliverable scope.
What if a fine-tuned model underperforms your baseline?
Acceptance criteria are written into the SOW with quantitative thresholds on the Inspect-AI suite. If a run misses the threshold, we iterate within the agreed budget, or we refund the milestone — at your option.
Data & security
Do you use customer data to train your own models?
No. Customer data is processed strictly under the contract scope, never used for general model training, and deleted on request at the end of the engagement (see Privacy).
Can we self-host everything you deliver?
Yes — that is the default. Every artifact runs on your infrastructure (cloud, dedicated, or on-premise). We hold no production keys after handover.
How are releases secured?
Every release artifact is signed with Ed25519, and every build event is hash-chained into a Merkle log root. Detailed in our Security Policy.
Engagement
How long does an engagement typically last?
2 to 12 weeks depending on scope. Most MCP server engagements are 2-6 weeks, fine-tuning 3-8 weeks, multi-agent swarms 6-12+ weeks.
Can you start next week?
Usually yes, if scope is clear and the SOW can be signed within 48-72 hours of a discovery call. Slots are limited — book at cal.com/julien-compain.
Do you offer ongoing support after handover?
On request, we offer a structured support agreement (monthly retainer, response-time SLA, quarterly review). It is opt-in and never bundled into the initial engagement.
Technical
Which models do you fine-tune?
Open-weight models: Llama 4, Qwen 3.7, DeepSeek V4, Kimi K2.6, GLM 5.1, Gemma 4. Frontier models when distillation is the goal: Claude Opus 4.8.
Which fine-tuning methods?
QLoRA, GRPO (DeepSeek-pioneered reasoning method), DPO, KTO, distillation, full fine-tuning when justified. Powered by Unsloth Studio and torchtune.
What MCP clients do you support?
Claude Desktop, Cursor, Windsurf, Cline, and any custom headless orchestrator that speaks the Model Context Protocol spec.
Compliance
Are you EU AI Act compliant?
We perform a risk-based assessment per engagement and design deliverables to meet the Act's requirements for the relevant risk category. Detailed in our Security Policy.
Do you support SOC 2 / ISO 27001 / HIPAA engagements?
Yes, as compliance adders to a base engagement. Expect +€15 000 to +€80 000 depending on the framework.