What I build at Aalee: Assist

Aalee Assist is the product I spend my days on at Aalee AI. It is a no-code platform for creating, training and deploying AI assistants that answer from a company's own knowledge, and it is where most of my engineering work goes.

The problem it solves is a familiar one. Customer conversations arrive on a website chat, WhatsApp, Telegram, Instagram, Messenger, Slack, email and Mattermost, each with its own inbox and its own history. Bolt-on chatbots answer what they recognise and dead-end the rest, so the handover to a person comes late and without context.

Assist puts all of those channels behind one assistant and one shared dashboard:

• One assistant, eight channels. The website widget plus seven messaging channels, all in a single thread view.
• Escalation decided by intent. Through tool calling, the assistant decides when to answer, when to hand over, and which agent should take the conversation, with the full history carried across.
• Bring your own key. Teams connect their own accounts across 15+ LLM providers and switch models without a migration.
• Grounded in your knowledge. Assistants are trained on the company's own content rather than answering from general knowledge.

I lead Assist and built most of it, including the LLM and tool-calling layer, provider routing, the embeddable web widget and the agent dashboard. The stack is Python and FastAPI on the backend, React and TypeScript on the front, with Vertex AI, Redis and WebSockets underneath.

You can try it at assist.aalee.ai↗.