FORMALIT

We generally do not speak of our customers in public. Below are some recent customer projects explained from a high level or personal projects that can be explained in detail. For more details also check me at LinkedIn

AI Driven development

In this initiative we developed several applications using ChatGPT with an estimated speed improvement of a factor around 2-5.

One of the applications was a full stack application with React front-end, javascript API, and postgres database, all hosted in Azure Cloud as static web app, web app and postgres database, respectively. Automatic reqression tests for UI and API ensured correctness. An overall approach for how to run a project and develop the application in steps that addresses the main known reservations for AI driven development is included.

This paper explains the approach.

This presentation to faculty of the IT-University of Copenhagen adds some history.

This initiative offers the approach to customers.

The UI and API code for one of the generated applications is available at GitHub.

Cloud Migration

In this customer project we were responsible for migration of a customer management system handling a million customers.

Both the new system and the migration application ran on AWS.

The total duration of the most critical step went from 9 days in the initial version down to just over 1 hour. The optimizations included tuning as well as both horizontal and vertical scaling.

This paper distills the optimizations into general techniques that can be used by most migrations.

Configurable Multi-tenant Application for Global Enterprise

In this customer project we developed a sophisticated application to be used by departments in many different countries of a global enterprise. The requirements from different divisions varied significantly, and the challenge was to develop a product that would cover all departments in all countries without knowing all the requirements in advance. The product ended up being requested even from external customers.

This paper explains the approach.

Wordle Algorithm

In this personal project we developed a Python program that can guess any Wordle secret in at most 7 guessses. It is known that by manual tinkering, it can be reduced to 6.

The algorithm is based on a well-known ingenious approach from the 1970s to solve Master Mind puzzles.

The code was heavily optized from the initial version to the final code, with a factor of around 1.000.000. This paper and presentation explain the algorithm and optimizations. The code is available at github.