Case study
Crypto AI Assistant.
Three AI models — OpenAI, Gemini and a sentiment model — integrated inside one live Flutter crypto app, delivered and rated 5.0 by the client.
- Engagement
- Fixed-price AI integration, Sep 2025
- Role
- One senior engineer — integration and delivery
- Platforms
- Flutter, iOS and Android
- Outcome
- Three models shipped, closed at 5.0
The problem
The client had a working Flutter crypto app and wanted AI inside it — not a chatbot bolted to the side, but assistance that belonged to the product. The hard part of that request is rarely the API call. It is deciding which model answers which question, what happens when one is slow or down, and how the app behaves when a model returns something confident and wrong.
What we built
Three models integrated behind one interface: OpenAI and Gemini for the assistant, and a sentiment model reading market tone. Routing sits in the app rather than in the prompts, so a model can be swapped or failed over without touching the screens that use it.
The assistant itself is a chat surface inside the existing app, talking to the product’s REST APIs so it answers from the user’s actual data rather than from the model’s memory. Sentiment analysis runs alongside it, which is the feature that most needed a confidence floor — a market-tone reading presented too certainly is worse than no reading at all.
The result
All three models shipped into the live app and the contract closed at 5.0.
“Arslan did an outstanding job on my project. He successfully integrated multiple AI models (Gemini, ChatGPT, and an additional model) into my Flutter crypto app…”
Stack
Flutter · Dart · OpenAI · Gemini · Sentiment analysis · AI chat assistant · REST APIs