Axion Ray
Data Analyst, AI and Advanced Analytics
AI quality-intelligence software for manufacturers.
Diebold Nixdorf ATMs
- Flagged ATM failures before a technician raised them by combining IoT telemetry with LLM-tagged technician notes, on a data model I owned across service, telemetry and transaction data.
- Showed that over 80% of platform activity sat with 3 users through a weekly usage report (SQL over ClickHouse, delivered to Slack) that set the account team's adoption plan.
Nexperia Semiconductors
- Traced field failures back to the originating wafer with an LLM tagging model over technician notes, cross-validated against source records. The client used it to fix the faulty wafer source.
Eaton Power management
- Built one lifecycle view from 8 source datasets (purchase to warranty claim to replacement), then tagged failure, root cause and fix with LLMs to isolate recurring failure modes.
QSC Audio
- Turned customer reviews into queryable failure and sentiment data using GPT-5.6 with a structured JSON schema, validated by sampled review and an LLM-as-judge pass.
Across all accounts
- Cut LLM tagging cost by 60% by deduplicating near-identical records, batching inference calls and stripping boilerplate before tagging.
- Cut metric validation from 3 hours to 30 minutes with an internal Metric QA tool that pinpoints whether a broken metric starts in SQL logic, source data or the pipeline. Adopted by 3 analytics teams.