The Global Alliance for Genomics and Health (GA4GH) participated in the Google Summer of Code for the eleventh year. Google Summer of Code (GSoC) is a global programme that provides new developers with an introduction to open source software development. This year, GA4GH acted as a mentor organisation that supervised four project teams.
Reflecting on the importance of GA4GH’s role in mentoring participants in this programme, GA4GH Tech Team Lead Jimmy Payyappilly said, “GA4GH is open source, which means all the technical artefacts produced under GA4GH can be used by anyone around the world. Through participation in Google Summer of Code (GSoC), GA4GH has ignited interest in graduates about the increasing role of technology in genomics and its importance in standardisation and interoperability. GA4GH continues to enable and empower these students to contribute to the field of genomics, as part of their courses as well as future community members.”
One of the programme mentors, Alex Kanitz, added, “Since 2019, the ELIXIR Cloud & AAI Driver Project has relied on GA4GH's GSoC participation to recruit over 15 incredible open-source contributors into the life sciences. Mentoring these students as they build critical ecosystem tooling — spanning client and server-side implementations, middleware, and compliance testing — is an extremely valuable way to promote open standards. The fact that some of these contributors are still working in the field five years later speaks volumes about the programme's lasting impact.”
GSoC programme participants presented their projects in a webinar on 28 October. Learn more about the GSoC projects through summaries and reflections shared by the participants themselves.
GA4GH-RegBot: AI-Assisted Regulatory Navigation for Genomic Data Sharing
Programme participant: Ruiyang Chen (New York University — Undergraduate)
RegBot helps researchers navigate GA4GH policies and jurisdiction-specific regulations. It retrieves relevant provisions from a curated regulatory corpus and provides citation-grounded guidance for human review rather than legal decisions.
“This project taught me that trustworthy AI requires not only model performance, but also traceable evidence, reproducibility, and clear boundaries for human oversight. I gained practical experience in building a primary-source corpus, hybrid retrieval, citation grounding, evaluation, full-stack development, and deployment.” — Ruiyang Chen
Workflow Clinic: Automated Diagnosis and Remediation Engine for Cloud-Ready Scientific Workflows
Programme participant: Revaa Rathore (Polaris School of Technology, Bengaluru — Undergraduate in Computer Science, specialising in Artificial Intelligence & Machine Learning)
Moving complex genomics workflows (Nextflow/Snakemake) to cloud infrastructure often fails due to portability blockers like unpinned containers, missing compute limits, hardcoded paths, or leaked credentials. Workflow Clinic solves this with a two-phase engine:
- Diagnostic and AI Critic Engine: combines sub-5ms offline Abstract Syntax Tree (AST) static analysis with a LiteLLM-grounded AI Critic to audit workflows for compliance with GA4GH cloud standards.
- Workflow Doctor Remediation Engine: automatically generates and applies multi-tier patches (AST directive injection, scoped regex parameterisation, and LLM refactoring) validated against compiler syntax checks (verify_fix()) and publishes atomic pull requests directly to remote GitHub repositories.
“Participating in Google Summer of Code with GA4GH has been a truly transformative experience. Coming from a web development background, this project pushed me deep into compiler design, AST parsing, bioinformatics workflow portability, and grounded AI orchestration. Across 14 weeks and 40+ pull requests, testing against real-world production pipelines like nf-core/fetchngs taught me how crucial it is to design for real-world complexity rather than idealised assumptions. Working closely with my mentors, Alex Kanitz and Javed Habib, fundamentally reshaped my engineering mindset. Their guidance to always 'prioritise correctness over cleverness' and maintain rigorous testing discipline (240+ tests, 0 type errors) showed me what it truly takes to build robust, trustworthy infrastructure for open-source science. I am deeply grateful to GA4GH, ELIXIR, and my mentors for this incredible journey!” — Revaa Rathore
GA4GH-Aligned Trustworthy Federated AI for Genomic Ancestry Prediction
Programme participant: Vidit Khandelwal (Indian Institute of Technology, Varanasi, India)
The project demonstrates a cloud-native, privacy-preserving federated learning proof-of-concept for genomic ancestry classification aligned natively with GA4GH standards. It implements a stateless, batch-driven architecture using GA4GH Data Repository Service (DRS) for data resolution and Task Execution Service (TES via Funnel) on Kubernetes to enable FL model training across isolated institutional silos.
“GSoC with GA4GH was an incredible learning experience for me, giving me the opportunity to work on a meaningful real-world problem while learning how to design and build a federated learning system from scratch. Beyond the technical skills, I really valued the discussions, mentorship, and came away with a much better understanding of how to build software that can be useful and reproducible beyond my own machine!” — Vidit Khandelwal
Decoupled Standalone/service-info microservice.
Programme participant: Gopi Kishan (Polaris School of Technology)
The project aimed to design a standalone Go microservice that runs in its own separate Kubernetes pod. It is configured to intercept and serve standard/service-infometadata queries locally, while redirecting all other standard application traffic to its respective backend application.
"Participating in GSoC 2026 has been an incredibly rewarding journey. Designing this standalone microservice to run in a separate pod and manage traffic routing taught me how to tackle real-world, cloud-native architecture challenges. I am deeply grateful to my mentors Pavel, Alex, and Javed for their constant support, and I am excited to stay involved and continue contributing to the GA4GH ecosystem!" - Gopi Kishan
