TT Collective

Copilots

Infra Crew

Jane

Nine-Nine

Product, Design & Impact

FMCG & cie

socraft

Undefined

DONOMA

Julia Creative Studio

Agence Dentifrice

SASWAVE

Les experts CTO / CPO

Inexplo

Kadabra Studio

FineStack.io

Sypher

LOT Systems

Issue: The client is operating an interface for a physical product subscription service, with a user base distributed around the globe. They wanted to implement a journaling feature to the interface, which based on the users answers creates a digital clone in the backend.Solution: The system was build with python, LangChain, chroma (vector db) and the openai API. It was deployed on multi-regional Cloud Functions and Cloud Run services on Google Cloud Platform (GCP). A dedicated service API was designed and used to integrate with legacy backend services as well as their Vue.js frontend. We also developed the Vue.js integration of the API. The functionality of the service consisted of features such as: <ul><li>user journal upload and vectorisation</li><li>user behavioural model creation </li><li>integration of ambient data like location, time, temperature, weather, etc</li><li>inference based on user model, journal entry, ambient data, for user interaction design (in the form of question asked to the users with the goal to build more precise user models related to their product preferences and subscriptions)</li></ul>Duration: Two months.Impact: The client was able to launch a cutting edge LLM RAG system with custom functionality and roll-out to their users.

Development & deployment of LLM RAG system

Google Cloud Platform

Python

DevOps

LLM RAG on GCP for a product subscription service

Univention

Issue: The client was facing a tremendeous loss in productivity due to the lack of automation -- tests were carried out by hand, software packaged and shipped with various scripts which only a few people were able to correctly execute. The production system was in an unmaintainable state, as no one really knew what exactly was lately altered by whom.Solution: As the project was already hosted in a GitLab repository, we firstly developed a pipeline. Easily, we were able to automize code linting, unit tests, setting up test systems in order to carry out integrational tests, packaging and providing these packages. In a next step, we developed a test-harness where the team could easily integrate tests on various levels, e.g. unit tests written by developers as well as end-to-end tests by consultants.Furthermore, we carefully started to carry out the deployment process and administration of systems using Ansible in order to make deployments accessible, less error prone and have a single point of truth of the state of the system.In a next step, we also incorporated the Ansible deployment into the CI-/CD-pipeline, such that the customer was able to roll out new versions with a single click.The team is now in a position to get releases out more quickly. At the same time, due to test automation it is more confident of the quality and correctnes of the product. Furthermore, due to automation more resources are available to work on tasks that really matter.

Orchestrate test automation, building packages, test deployments and deployment to production in a pipeline in order to stream line the entire development cycle. 

Consulting IT & Cyber

Developing a CI-/CD-Pipeline in GitLab

Software Architect and System Engineer

Site Reliability Engineer

Depending on your individual requirements we plan and put in action CI/CD pipelines to make deployments as easy as a single click.

Plan and implement CI/CD pipelines

We bring your infrastructure to the next level, applying DevOps best practices.

We help you find the architecture for your business that suites you just right. We also help finding your requirements by studying your business case.

System Architecture

758 collectifs