About

I build production data platforms and AI infrastructure. Over 8+ years at Roku, Groupon, AtkinsRéalis and Eatclub, my work has spanned recoverable ingestion, cloud migrations, shared Iceberg tooling and governed data access.

At Roku, I’m a Senior Software Development Engineer on the Ads Data Platform, leading four engineers informally. I work on a 70+ PB lake estate, with ingestion processing 250M+ events daily and a maintenance framework covering 500+ mission-critical tables. Separately, my Hive-to-Iceberg migration platform has migrated 450 tables with no production incidents.

Before Roku, I migrated 2,500 Spark jobs from on-premises infrastructure to GCP Dataproc at Groupon, alongside a separate AWS Keyspaces to GCP Bigtable datastore migration. At Atkins, I built image and GIS pipelines on Databricks.

The through-line in most of what I build is governance that does not rely on people remembering to be careful. A policy written in a wiki is a suggestion; a policy enforced in the query path is a guarantee. Most of my recent work — sensitive-data classification, certified-metric enforcement, approval-gated migrations, identifier rotation for deletion — is some version of moving a rule from documentation into the critical path.

I am also interested in where language models genuinely help with this and where they do not. They are good at assembling context and drafting structure, and they are not accountable for a corrupted production table or a misclassified PII column. The systems I have had most success with lean hard on the first and keep a human firmly attached to the second.

Elsewhere

I am on GitHub andLinkedIn, and reachable atshril.iitdhn@gmail.com.