Description
Summary We've built an AI SaaS product. We're not convinced it's genuinely AI-first. We want an experienced AI engineer to audit it honestly and tell us what needs to change. About us We're a UK startup building AI software for professional-services SMEs, starting with architecture practices. We have a working product, 105+ practices on our waitlist, and industry endorsements. Our current build was delivered by an external development team. The problem: it was specified feature-first, not AI-first. We suspect some of the "intelligence" is thinner than the product promises. We'd rather find that out now than after launch. What we need — the audit A focused technical review of the existing system, delivered as a written report. Specifically: Codebase & architecture review — how the application is structured and where that helps or blocks an AI-first approach AI layer assessment — prompts, model calls, retrieval, grounding, evaluation. What's genuinely AI-native vs. cosmetic? Data model review — is the data structured in a way that supports real retrieval and reasoning, or does it need re-architecting? Gap analysis — compare what the product currently does against our functional specs and product promises Recommended target architecture — a concrete AI-first architecture with a prioritised build sequence Risk flags — hallucination exposure, multi-tenant data isolation, cost/latency issues, anything that would fail investor or customer diligence Deliverables Written audit report (10–20 pages or equivalent) Recommended AI-first target architecture, with a phased build sequence A 60-minute walkthrough call with the founders Our current stack Node / TypeScript · React · PostgreSQL · Claude (Anthropic) API · Microsoft Graph · Azure We're stack-pragmatic, not stack-precious. If the audit concludes something should change, we want to hear it. Who we're looking for You've shipped real AI features to production — not demos or prototypes Hands-on with RAG, vector retrieval, LLM orchestration and evaluation Experience with multi-tenant SaaS and the data-isolation problems that come with it Comfortable reading someone else's codebase and giving a direct, unvarnished assessment Clear written English — the report is the deliverable Bonus: Anthropic/Claude API experience, or vertical/domain-specific AI Timeline & budget Fixed price, 2–4 weeks, part-time. Budget is stated above and negotiable for the right person. We'll provide repo access, functional specs and architecture documentation under NDA. Potential for ongoing work If the audit goes well, there's scope for a longer-term role leading the rebuild. That would be an equity-based, part-time arrangement initially, with salary following post-launch and post-funding. We're being upfront about this: the audit is the paid engagement — anything beyond it is a conversation, not a promise.