Technology behind the adigi suite

The adigi suite consists of multiple products individually composed for our partners, the travel companies:

  • HolidayOffer focuses on automating the creation of individual holiday offers and the conversations around.

  • The BeratungsBoxen (AdviceBoxes) provide inquiry ingestion forms for high-traffic partner websites.

  • HolidayService automatically answers service inquiries of our partners’ customers.

  • HoLUnder is an NLU solution specifically for the travel industry and the terminology used therein.

  • The Message Router distributes incoming inquiries from different channels to the best matching products and other CRM solutions for travel companies.

+++ Join our team: We’re hiring! +++

Most products are developed using Ruby as programming language and Ruby on Rails as primary framework. Data resists in MySQL databases and the AWS S3 object store. Git is ubiquitous nowadays.

We strongly focus on Behaviour Driven Development with Cucumber and RSpec and have a test coverage of over 98 % in all our products. BitBucket assists our agile methology with eg. code reviews and concurrent feature branches.

Our team primarily consists of DevOps, therefore our infrastructure is defined as code. We heavily use Docker and Terraform to approach this. adigi is completely “in the cloud” with Amazon Web Services to focus on domain instead of infrastructure challenges.

AWS services of different abstraction levels are applied, eg. ECS/Fargate, ECR, Comprehend, CloudWatch, ELB, RDS, Route 53, SES, SNS, SQS.

adigi takes security, privacy protection, and the GDPR very serious and encrypts all data with KMS.

HoLUnder is a Python-based NLU stack. The Team leverages PyTorch and for example transformers and spaCy enriched with custom strategies to solve NLP tasks like entity recognition and classification. Development tools like Jupyter Notebooks and again Docker are a natural fit. We went serverless with AWS Lambda for unsophisticated Terraform-based deployment and seamless scaling. AWS EC2 instances are utilized for the training of deep learning models (e. g. BERT).

The BeratungsBoxen consist of high-performant JavaScript in combination with CloudFront as caching CDN. The forms are embedded in a strongly isolated way to decouple them from the partner’s website and their specific UI and performance needs.

Interested in more infos?

Contact CTO Florian Dütsch (E-Mail, +49 9602 37080-91)

2nd place at Travel Technology Award 2021 🏆!