Neeraj Sehgal

I built an automated solution to migrate 70+ Spring Boot microservices from OpenShift to GCP within stringent deadlines. The manual migration involved cloning each service, upgrading the libraries, configuring GCP changes, replacing deprecated code, fixing the test cases, and pushing the code to Git. The automated solution codified the cloning, configuring, and upgrading of seven library paths from documentation into a Python back end and created GitHub PRs across the full estate simultaneously.

I integrated an AI diagnostic engine with a validation harness, applied test fixes to a temporary branch, and executed the full test suite, committing changes only if the pass rate genuinely improved.

As a result, I reduced per-service effort from four days of manual work to 10 minutes for all 72, simultaneously demonstrating a 70% reduction in effort.

The same solution will continue to be used for future library upgrades, saving 70 person-days of effort whenever the next library upgrade is needed, approximately every three months.

A conversational UI was rolled out to IKEA co-workers to assist with querying Confluence content and company policies in plain, simple English.

I designed and delivered a RAG pipeline for the content assistant, with full document ingestion of Confluence pages and policy documents, chunking, embedding via Azure OpenAI Text Embedding model, and indexing in Azure AI Search with vector retrieval.

I also identified and resolved a near-duplicate document problem pre-ingestion using a cosine similarity check, preventing vector index pollution and resource wastage. I evaluated the retrieval output using RAGAs metrics, Azure AI default evaluation framework with distributed traces, and a curated golden dataset scored against actual responses.