I left a welfare support job to build an AI platform for the communities I used to serve — here's what 6 months of solo founding looks like

I spent years working as a Welfare Support Assistant at a homeless hostel in Bristol. Every day I'd watch people struggle to find the most basic local services — free meals, emergency support, mental health resources. The information existed, but it was scattered across dozens of websites, PDF leaflets, and word of mouth.
That frustration became the seed for what I'm building now.
I'm Ibrahim, solo founder of Nyata AI — a conversational AI platform that connects people to local community services. Think of it as a neighbourhood companion that actually knows what's available near you and can help you access it through a simple chat.
What I've learned so far building this:
The hardest part hasn't been the tech. It's been figuring out which problem to solve first. When you're building for communities, everyone has an opinion on what you should prioritise. I've built taxi booking flows, beauty marketplace features, charity directories — some of it stuck, most of it taught me what not to build next.
A few honest takeaways from the journey:

1. Your background is your unfair advantage, not your CV.
   My welfare work wasn't "tech experience" on paper. But understanding how a rough sleeper actually navigates services? That's something no amount of user research could replicate. If you're coming from a non-traditional background, don't discount it.
2. Bristol has been the perfect testbed.
   Building for a specific city first (rather than "everyone, everywhere") forced real constraints that improved the product. I know the charities, the council services, the gaps. Going hyper-local before scaling is underrated.
3. Fine-tuning an AI model on local knowledge is harder than it sounds.
   I'm running a fine-tuned GPT model trained on Bristol-specific service data. The challenge isn't the API call — it's maintaining accurate, up-to-date information about organisations that change their hours, close down, or move. The data layer is the real product.
4. Onboarding your first non-technical partner is a milestone.
   I just brought on a Business Development Partner. Writing that first formal agreement, setting up a company email, creating an onboarding pack — it made the whole thing feel real in a way that shipping code never did.
5. Trade marks, ICO registration, App Store compliance — the boring stuff matters.
   Nobody talks about this, but the admin side of building a legit company in the UK takes serious time. ICO registration, data protection policies, App Store review guidelines, trade mark applications — each one is a mini-project.
   Where I'm at now:
   Live on iOS and Android. Fine-tuned AI model. One partner onboarded. Preparing to speak at a cybersecurity and inclusion conference at my old university later this month. Exploring Innovate UK funding.
   Revenue: £0. Users: growing slowly. Lessons: countless.
   Happy to answer questions about building AI for local communities, navigating UK startup admin as a solo founder, or the realities of going from welfare work to tech.