ViewsML

Cloud AI, Infrastructure Modernization 

Overview

ViewsML is transforming spatial biology for translational research and precision medicine using AI deep learning. By virtualizing immunohistochemistry (IHC) and biomarkers – and thus bypassing this wet lab process – they give researchers back 100% of their IHC time, 50% of their budget and richer data sets. 

Challenges

Celestial's Solution

Celestial built a Full stack application which uses Biopsy report as an input and produces a Deep zoomed image(DZI) as output using ViewsML’s algorithms. Using this technology makes detecting cancer cells easier for pathologists. While the point above was a 100-foot-high level view of the application we built for ViewsML, there was a major challenge that our Devops team was able to solve.

We have worked on all six primary aspects of the ViewsML Application comprising UI, FrontEnd, BackEnd, DataBase, DevOps, and also on their BlackBoxed ML piece as well. 

DevOps Work 

We orchestrated a comprehensive transformation of ViewsML’s infrastructure leveraging DevOps methodologies. This involved containerizing the entire infrastructure on Azure using Docker, implementing multi-tenancy through Kubernetes-managed clusters, and establishing robust CI/CD pipelines via GitHub Actions across development, QA/Staging, and Production environments. Automated deployments were facilitated using ArgoCD, while monitoring capabilities were enhanced through Grafana and Prometheus. 

Application Work (FE and BE)

FrontEnd and UI 

Recognizing the need for a more intuitive user experience, we undertook a complete overhaul of ViewsML’s dashboard UI. This involved redesigning the UI flow and introducing new sections for tasks such as image uploading, selection of algorithms, image processing, and visualization. Additionally, advanced tooling for the DZI Image Viewer was incorporated, enabling users to perform annotations, measurements, and more.

Redefined Application Flow

Back End

Our team diligently worked on achieving multi-tenancy and refining the application flow within ViewsML. The BackEnd infrastructure was developed using GoLang, with a focus on optimizing performance and scalability. Significant efforts were also directed towards enhancing the Algo Container, the core component of ViewsML, to support multi-tenancy efficiently, leveraging Python. 

Below is the list of new APIs that were created: 

POST {tenant_url/oauth/authorize} Auth
GET {tenant_url}/api/1.0.0user_uploads/{user_id} List User Uploads
POST {tenant_url}/api/1.0.0user_upload}s Upload File
GET {tenant_url}/api/1.0.0/.well-known/config.json} Tenant Config
POST {tenant_url}/api/1.0.0/cancer_sample_uploads} Processing sample,Send request Algo
GET {tenant_url}/api/1.0.0/proccessed_samples} {tenant_url}/api/1.0.0/cancer_sample_uploads}
GET {master_url}}/api/1.0.0/master/clients} Master Tenant list all tenants
POST {master_url}}/api/1.0.0/master/clients} Creat Tenant
POST {master_url}}/api/1.0.0/master/clients/{tenan_id}/tenantstatus} Tenant Status

Database

Utilized PostgreSQL to cater to ViewsML’s evolving needs. This involved defining and implementing new data models to accommodate the enhancements and changes introduced throughout the project lifecycle. 

Technology Used :

Outcomes

Wondering what Celestial has to offer?

Celestial respects your privacy. No spam!

Thank you!

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Stay up to date with Celestial