aineo
An AI application for a Swiss AI company, started from the first commit and led through design and build.
- Role
- Lead Engineer
- Stack
- 1 × React
- 1 × Python
- 1 × LangChain
- 1 × OpenAI
The job
aineo AG is a Swiss AI company. It reads documents, transcribes conversations and builds AI workflows for businesses that need their data handled under Swiss data protection law and kept in Switzerland.
I started the application at Bizdesire and led it from there: the product design, the front end and the shape of the backend. Nothing existed before the first commit.
Two halves, two languages
The front end is React. The backend is Python with LangChain, which is where the AI work belongs: the Python ecosystem is where the model tooling lives, and LangChain gives the calls to OpenAI a consistent structure instead of one-off request code. The interface stays a client of that backend and never talks to the model directly.
Answers that read like documents
A model's answer is rarely a single paragraph. It comes back as Markdown, with headings, numbered steps, tables and code, and shown as raw text it is a wall of symbols. The front end renders that Markdown properly, so a long answer reads like a formatted document rather than a transcript.
A theme that is not hard-coded
The interface was built on a dynamic theme rather than colours fixed in the stylesheet, so the look of the application could change without rewriting its components. I designed the product and built that theming into it from the start, instead of adding it later on top of fixed styles.
Questions
- What is aineo?
- aineo AG is a Swiss AI company that reads documents, transcribes conversations and builds custom AI workflows for businesses, with data kept in Switzerland.
- What did Abhishek Thakur build for aineo?
- He started the application and led it at Bizdesire: the product design, the React front end with dynamic theming and Markdown rendering of OpenAI responses, and the Python and LangChain backend behind it.
- Why render AI responses as Markdown?
- Language models answer in Markdown: headings, lists, tables and code. Rendering that structure properly makes a long answer readable, where showing it as raw text leaves symbols scattered through the reply.
Last updated