StartGo Mobility's original AI platform initiative — secondary to our core ride-services business.
ONE AI SPARK is our in-house effort to build an AI assistant platform and, over the long term, our own original AI models — developed independently rather than as a permanent wrapper around another company's AI. It includes a chat assistant application, a backend gateway, and an active research program into original model architecture, tokenizers, and training data.
Implemented and working, as of this page's publication.
A Flutter-based chat application with conversation history, built with an interface-first design so backends can be swapped without UI changes.
A Node.js gateway handling authentication, rate limiting, and request routing between the app and any approved AI backend.
A dedicated coding-assistant screen (generate, explain, debug, refactor, generate tests) with full provider/error handling — currently honestly reporting "no backend configured" until a qualified backend is approved.
A from-scratch Transformer architecture, tokenizer, and training pipeline, developed and tested entirely in-house — currently a small-scale research checkpoint, not a production model.
Ongoing research into code-aware tokenization and a provenance-controlled dataset pipeline for future coding-model training.
Every capability not yet backed by a real, approved AI model says so explicitly in the app, rather than faking a response.
Clearly distinguished from what's implemented above.
StartGo Mobility