Add FedViz link to BOSC blog post

Author: curious7-web

content/posts/2026-08-05-aditya-karna-fedviz-bosc.md

@@ -26,7 +26,7 @@ Attending BOSC/ISMB 2026 in Washington, D.C with support from the Open Bioinform
 
 Presenting FedViz at BOSC/ISMB 2026 in Washington, D.C.
 
-I came to BOSC to present FedViz, a project I worked on around federated biomedical data. The basic idea behind FedViz is simple: before different sites try to train models together using federated learning, we should first ask whether their data is actually ready to be used together. Different biobanks and cohorts may collect different variables, describe them differently, or have missing information in different places. If we ignore that stage and jump directly to modeling, we may build systems that look technically impressive but are not actually reliable.
+I came to BOSC to present [FedViz](https://github.com/collaborativebioinformatics/FedViz), a project I worked on around federated biomedical data. The basic idea behind FedViz is simple: before different sites try to train models together using federated learning, we should first ask whether their data is actually ready to be used together. Different biobanks and cohorts may collect different variables, describe them differently, or have missing information in different places. If we ignore that stage and jump directly to modeling, we may build systems that look technically impressive but are not actually reliable.
 
 In our work, we looked at metadata across 14 international cohorts and indexed 11,511 unique variables. Out of those, only 120 variables appeared in at least eight cohorts, which gave an immediate model-ready surface of about 1.04%. That number stayed with me because it made the problem very clear. The hard part is not only training a model. A lot of difficulty comes earlier, from understanding what is available, what is missing, and what can realistically be compared across sites.