Shivanand KumarBengaluru, India
IIT PatnaM.Tech · AI & Data Science EngineeringGenerative AI · Agentic AI · Databricks · Spark
I build reliable data and AI systems—from Databricks and Spark platforms to GenAI automation, agentic workflows, retrieval and evaluation.
Selected production impact
HP · 2024 — Present · Public-safe summary
A public-safe view of engineering ownership across data platforms and Applied AI. Employer code, data and confidential architecture remain private.
Built a graph-orchestrated workflow across 40 enterprise Databricks jobs to turn failure evidence into consistent, reviewable analysis.
Automated configurations, schemas, DDLs and 20+ transformations across a governed data-product workflow.
Improved monitoring, migration and distributed-processing workflows across high-volume enterprise data platforms.
Independent public builds
Three reproducible projects built with synthetic data, versioned evaluation and clear scope boundaries. Each case study links to output, source, tests, architecture and disclosed limitations.
A tool-using reliability agent with an optional OpenAI planner that investigates synthetic pipeline incidents and proposes, but never executes, remediation.
Measures data scarcity, generates controlled document variations and compares 1x, 10x and 100x scaling on a separately versioned synthetic holdout with disjoint IDs.
A fully synthetic Python lab that separates ordinary functional correctness from six independently observed runtime behavior constraints.
These are independent public implementations, not employer systems. They contain no employer code, data, configuration or confidential architecture.
The learning notebook
Focused experiments, working notes and what still needs testing.
Experience and approach
I work where data systems meet Applied AI: reliable inputs, measurable evaluation, controlled actions and clear operator handoffs.
Mechanical engineering shaped my systems mindset. I carried it into data platforms and now into production-focused Generative and Agentic AI—building dependable foundations and evaluation-driven, human-accountable AI workflows.
Databricks platforms, production automation, observability and Applied AI workflows
SQL pipelines, analytics automation and LLM data and evaluation work · director-level recognition for automation impact
16-member cross-functional operations team and SQL-led operating reviews
Led a 16-member operations team at BYJU'S and now bring the same emphasis on accountability, measurable outcomes and clear handoffs to engineering work.
Professional contact
Applied AI, GenAI, Agentic AI and Data & AI engineering.
