Claude Code developer — AI platform build (UK only

Hi,

Thanks for sharing the project brief. It looks like a genuinely interesting platform, and I can see a lot of thought has gone into the structure, data sources and session-by-session build plan.

That said, I think the current budget is significantly below what would be realistic for the scope described.

Even with Claude Code being used as the primary build tool, this is not a simple “AI-assisted build”. The project includes PostGIS data ingestion, Companies House and FSA data processing, entity resolution, Google Places/Maps API integration, scoring logic, solar detection, website scraping, AI-generated outreach hooks, audit trails, validation workflows, a FastAPI backend, and a React dashboard. That is a substantial commercial intelligence platform, not a small MVP.

I would also be concerned about the ongoing running costs. The brief mentions Google Places, Google Maps Static, OS data/API usage, website scraping, AI-generated outreach hooks and potentially recurring DataLedger/monthly delta processing. Depending on volume, those API calls and processing costs could become meaningful quite quickly, especially if the system is analysing large numbers of buildings or company records.

So my honest view is that the £600–£1,500 range is way off for a properly built and tested version of what is described. It may be possible to create a narrow prototype within that range, but not the full platform outlined in the brief.

A more realistic approach would be to define a smaller Phase 1 MVP, for example:

1. ingest a limited dataset
2. match companies using one or two resolution methods
3. create a basic scoring model
4. export ranked leads
5. leave aerial imagery, AI hooks, full validation workflow and dashboard polish for later phases

That would make the build more manageable and allow the real-world data/API costs to be measured before committing to the full system.

Happy to discuss a reduced MVP scope, but I wouldn’t want to pretend the full brief can be delivered properly within the current budget.

Best,
Jon

Hi Andy,

I’ve reviewed the brief in detail and just want to clarify a few points before applying:

Is prior hands-on use of Claude Code strictly required, or would strong experience with AI-assisted development workflows be considered?

How strict is the UK-based requirement? Would a UK-registered contractor working remotely be acceptable?

Are all APIs, datasets, and access credentials fully provisioned and ready for Session 1, or is any setup expected from the developer?

What level of autonomy is expected during each session versus client-guided execution?

For the PASS/FAIL verification, will you be providing automated checks, or is this manually reviewed?

If aligned, I’d be happy to proceed with a full application.

Hi Andy, would you be open to starting with Session 1 as a verified proof point before committing to the full eight-session build? That would let us work under the NDA, execute the first database/schema stage against your PASS/FAIL checklist, confirm the verification process and safe access route, then agree the remaining milestones once the first session has passed.

Is there a restriction on what models should be used? Is it up to the expert to pay for the Claude Code usage, or is that something that you will reimburse?