Skalujto.ai — websites, e-commerce and AI
What I see
Not a conference, not slides. Patterns that keep coming up in conversations with companies.
Sales reps copy from form to CRM, CRM to email. Zero qualification, zero context, zero history — while the customer waits for a reply.
Discovery, RFP, contract, build. By the time anything ships to production, it has already gone stale.
No baseline, no dashboard, no quality check. A few months in, nobody knows whether the solution earns money or burns the budget.
How I work
Every step has a fixed date and a concrete deliverable. You walk away from each with something that works — not another document.
Step 01 / 05
A conversation about what you have and what you want to achieve. You leave with a map: what to build, in what order, how much it costs and whether it makes sense at all.
Stack & tools
No framework religion. Frontend is usually Next.js and TypeScript. Backend is FastAPI and Python — an equal and large part of every project. For AI: local models on own servers or cloud when it makes sense.
FAQ
Audit: 3-5 days. Prototype (Next.js + optional FastAPI + AI): 2-4 weeks. Full production deployment: 6-12 weeks. Always fixed-date, fixed-scope — not hourly open-ended.
Fixed project price. You know the number before signing. Scope changes get a separate scope, quoted within 48h.
We pick for the problem, not the other way round. Frontend is usually Next.js and TypeScript. Backend is FastAPI and Python — an equal and large part of every project. For AI: Claude, OpenAI models, open-source models, RAG, local inference on own servers.
Yes. I run NVIDIA DGX Spark servers for local model inference. Client data doesn't have to leave their network — transcription, embeddings and inference can run fully on-premise. For cloud models (Claude, OpenAI) I always inform upfront which data goes where.
Always. Next.js + FastAPI repo on your GitHub, CI/CD on your Vercel or your own infra, keys on your accounts. No agency lock-in.
Metrics agreed before kickoff: conversion, time saved, Lighthouse score, AI answer quality (human-eval plus automated). Dashboard from day one of production.
Client data never leaves the infrastructure you control. Transcription runs locally (whisper.cpp on your machine). Embeddings live in pgvector on your database. API keys stay on your accounts. Encryption at rest (Postgres TDE) and in transit (TLS 1.3). If a model has to touch data outside the EU, we tell you before the contract, with the list of records affected.
OpenAPI is the contract between FastAPI and your system. Webhooks into Slack, Notion, Google Workspace, CRM (Pipedrive/HubSpot), ERP (Comarch/Symfonia) through official SDKs — never homebrew scrapers. If a system has no API, we write an adapter and document the limits. SSO (SAML/OIDC) where needed.
We don't build native mobile apps (React Native only if it completes a web project). We don't do classical ML (financial prediction, computer vision trained from scratch). We don't do enterprise infrastructure (multi-region Kubernetes, service mesh). We don't do long-tail SEO or ghostwriting. We say 'no' fast so you don't waste time.
30 minutes. You show me what you have and what you want to achieve. I tell you straight what can be built, how much it costs and whether it's worth it.