AI Software Development

Custom web platforms and AI features, built to be maintained

Internal tools, customer-facing platforms and machine-learning features that fit how your team already works, with clear access control and documentation your own engineers can pick up.

Plan a custom application

Indicative starting price: $49,999 (USD). Final price depends on scope.

Architecture of a custom web platformPeople use web interfaces. Their requests reach your application through an API. The application handles sign-in and permissions, business rules, an audit trail and background jobs, and stores information in a database and file storage. It connects to your existing systems such as ERP and email only through its API layer. A model service is optional. Automated tests, a deployment pipeline, logs and alerts, and backups apply across the whole system. People Staff app internal tools Customer portal self-service Admin console settings HTTPS API Your application Built and documented for your team Sign-in and permissions Rules and workflows Audit trail Background jobs Model service optional Data Database transactions File storage documents queries Your systems ERP Email Other APIs APIs Safeguards across the whole system Automated tests Deploy pipeline Logs and alerts Backups Architecture of a custom web platformPeople use web interfaces. Their requests reach your application through an API. The application handles sign-in and permissions, business rules, an audit trail and background jobs, and stores information in a database and file storage. It connects to your existing systems such as ERP and email only through its API layer. A model service is optional. Automated tests, a deployment pipeline, logs and alerts, and backups apply across the whole system. People Staff app internal tools Customer portal self-service Admin console settings HTTPS API Your application Built and documented for your team Sign-in and permissions Rules and workflows Audit trail Background jobs Model service optional queries APIs Data Database transactions File storage documents Your systems ERP Email Other APIs Safeguards across the whole system Automated tests Deploy pipeline Logs and alerts Backups
Users reach the application through a controlled API. The database is never exposed to clients, and your existing systems are connected through the application only.
Technical notes
  • Python and SQL are our core tools. The web framework is chosen to suit your team; we have used Django, FastAPI, React and Vue on past projects.
  • Permissions are enforced on the server, not only in the browser.
  • Database changes are versioned migrations applied by the deployment pipeline.
  • The model service sits behind the same API and access rules as everything else.

Where machine learning fits

Most of what a platform needs is ordinary software. We add a model only where a clear rule cannot do the job, and we test it on your own data first.

  1. 1
    Can a clear rule or a report solve it?For example, validation, routing, or a calculation.
    Yes. Build the rule. No model is needed.
  2. 2
    Is there usable data, or a set of documents to search?History to learn from, or content to retrieve from.
    No. Collect or label data first, or choose a different approach.
  3. 3
    Build the smallest approach and evaluate itTested on your data against a measure agreed in advance.
    Below the measure? Stop or change approach.
  4. 4
    Deploy behind the APILow-confidence cases go to a person.
    Same access rules as the rest of the application.
  5. 5
    Monitor and review mistakesRetrain, adjust or retire.
    Stopping is a valid outcome.

Problems this solves

  • Spreadsheets and email threads are being used as a system of record.
  • An existing application is hard to change and nobody is sure what will break.
  • You want to add search, extraction or prediction to a product without rebuilding it.
  • Several systems need to exchange data reliably through APIs.

What you receive

  • A working web application or internal platform, deployed to your environment
  • Role-based access and an audit trail where the process needs one
  • API integrations with the systems you already use
  • Automated tests and a short technical handover document
  • Machine-learning components where they earn their place, with accuracy measured on your own data

How the work runs

Each phase has an input from you and a written output from us. Ongoing support after launch is agreed separately.

PhaseWhat happensYou provideYou receive
Step 1Discovery and requirements Map the current process, the users, the data and the systems it touches. Agree what 'done' means. Time with the people who run the process; sample data or documents. A written scope with priorities.
Step 2Design Choose the architecture, data model and access rules. Decide where, if anywhere, a model helps. Decisions on hosting and security constraints. A short design document and interface sketches.
Step 3Build and integrate Develop in small increments with automated tests and connect to your systems through APIs. Feedback on each increment; test access to connected systems. Working software in a staging environment throughout.
Step 4Validate Test against your real workflows. For any model, measure accuracy on your own data. Review time from someone who knows the process. Written test results and known limitations.
Step 5Deploy and hand over Release to production, document operation and walk your team through it. Named owners for the system. The deployed system, documentation and code access as agreed.

Related work

Anonymised client projects. Client names are withheld because of confidentiality; the project page explains each one.

Client project, anonymisedKnowledge management

RAG Chatbot on Custom Knowledge Base

A retrieval-augmented chatbot that finds the relevant passages by vector search and answers from them, citing the source document.

Client project, anonymisedHealthcare

Healthcare Data Analytics Platform

An analytics platform that processes patient data to surface treatment patterns for review.

See all projects

From the blog

Practical articles from the engineers on this topic.

Sep 06, 2026

How I Actually Built a RAG Chatbot That Works in Production

Building a RAG chatbot that actually works in production means solving about 10 problems the tutorials skip — from hybrid retrieval and hallucination guardrails to conversation context and …

Mar 30, 2026

RAG Chatbots: Building AI That Knows Your Business

A retrieval-augmented chatbot is mostly a search problem with a language model on the end. Where these projects go wrong: chunking, retrieval quality, permissions, and having no way …

FAQ

Frequently asked questions

Do you build internal tools and customer portals for small businesses?

Yes. Typical projects replace a spreadsheet-and-email process with a web application that has roles, an audit trail and integrations. The scope is agreed in writing before work starts.

Can you build a chatbot that answers from our own documents?

Yes. We build retrieval-augmented (RAG) chatbots that answer only from your documents and cite the source, and we test them on real questions before launch.

Who maintains the software after launch?

You can maintain it with your own engineers using the documentation and tests we hand over, or agree ongoing support separately.

Can you add AI to an application we already have?

Often yes. We start by checking what data the application holds and whether a simpler rule-based approach would do the job before proposing a model.

What technology do you use?

Python and SQL are our core tools, with web frameworks and cloud services chosen to suit your team's skills and your hosting. We explain the choice and write it down.

How is the work priced?

The 'starting at' figure is indicative. After a short discovery conversation we share a written scope and an estimate based on what the project actually involves.

Who owns the code and the data?

This is agreed in writing before work starts. Our default is that you own what we build for you, and that your data stays in systems you control.

Planning an internal tool or customer platform?

Describe the process you want to replace or the product you want to extend. We will reply with questions and a suggested way to start.

Plan a custom applicationor email us