Hi @krisstern, @shivaylamba , and the rest of the Jenkins community,
I’m Jay Guwalani (GitHub: JayDS22), applying for the AI Chatbot to Guide User Workflow project for GSoC 2026.
Currently I’m a Research Scientist at the University of Maryland, working on large-scale LLMs and chatbot systems using Kubeflow, Jenkins, Airflow, and DevOps infrastructure (including the security angle). We’re targeting ICML and NeurIPS with this research. Before this, I spent 4+ years in industry building Jenkins CI/CD pipelines and production AI chatbots across three companies. So when I saw this project, it felt like it was written for my exact background.
Jenkins (5+ years across four orgs):
Started at Simplilearn, building Jenkins pipelines for reporting and Salesforce data extraction that fed a rule-based chatbot. At Bridgestone Mobility Solutions (3 years, promoted to Sr. Data Science Engineer), I owned the Jenkins pipeline architecture across their data platform: multi-stage builds, automated testing, environment promotion across dev/staging/prod, Terraform integration, and model deployment pipelines on SageMaker with MLflow. At Aya Healthcare, Jenkins CI/CD with Docker for multi-cloud deployments across OCI and AWS. And now at UMD, Jenkins is part of the MLOps infrastructure I work with daily for our LLM research.
I’ve spent years debugging failed builds at 2 AM, wrestling with plugin conflicts, and wishing the error messages would just tell me what actually went wrong. That frustration is exactly what this chatbot should solve, and I think having felt it firsthand matters when designing the solution.
Chatbots:
I’ve built multiple production chatbots over the years, starting with a rule-based one at Simplilearn. At Bridgestone: multi-agent architecture with LangChain, RAG pipeline over 15+ years of data, Claude/GPT-4o backbone, serving 1000+ concurrent users on Kubernetes. At Aya Healthcare: LangGraph with 5+ specialized agents, 2000+ concurrent users, voice interaction, and a responsible AI pipeline with toxicity detection. The architecture this Jenkins project needs is the same pattern I’ve built and refined across multiple production systems.
Why this project:
It’s greenfield, it’s Intermediate to Advanced, and the core problem goes beyond doc retrieval. It’s about understanding what a user is actually trying to do inside Jenkins and helping them get there. That needs someone who has built these AI systems and lived with Jenkins daily for years. I’ve done both.
I’ve set up my local Jenkins environment and I’m working on a proof-of-concept. I’ll follow up with technical direction and architecture details shortly.
Questions for the mentors:
-
Between workflow guidance (“How do I set up X?”) and troubleshooting (“Why did my build fail?”), which should take priority in the initial scope? My gut says troubleshooting has higher immediate impact, but I’d value your take.
-
For the in-Jenkins UI, would you be open to contextual inline help (hints on the build failure page, suggestions inside the pipeline editor) in addition to a sidebar chat panel?
-
I’m preparing my proposal doc. The draft feedback form seems to be closed. Is it okay to share the Google Doc link here for review?
Looking forward to your thoughts.
Jay Guwalani
Research Scientist, University of Maryland | MS in Data Science | LinkedIn | IEEE published, targeting ICML & NeurIPS