Software Engineer, Data & Platform at MidnaTech (2025-04 – Present)
Work across backend, data, and infrastructure on a two-product fleet platform: designing REST APIs and PostgreSQL schemas and implementing them in Python and SQL, planning work into written stories before coding, reviewing pull requests, and running an AI-assisted workflow against team guardrails I documented from bugs we actually hit.
- Built a serverless ELT pipeline that pulls Geotab Trips API telemetry into Supabase (PostgreSQL) and normalizes raw JSON logs into query-ready tables using in-database SQL transformations, with tests covering data integrity across the ingestion path.
- Designed and built the PostgreSQL tables and REST endpoints behind the customer dashboards, including 7/30/90-day rolling fleet metrics, and generated the frontend client types from the OpenAPI spec so the UI and API stay in sync.
- Maintain the deployment setup: Terraform for Railway and Supabase environments, GitHub Actions for CI/CD and gated migrations, and ~42 scheduled workers keeping fleet data current across seven production clusters.
- Added a narration feature using the Anthropic API that turns fleet metrics into plain-language summaries for customers, gated behind a license check; built React/Tailwind dashboards to surface the metrics.
Backend Data Developer at MidnaTech (2025-09 – 2026-04)
Backend Data Intern at MidnaTech (2025-04 – 2025-08)
FleetSense — Telemetry Ingestion & Fleet Analytics at MidnaTech (2025-04 – Present)
- Built a serverless ELT pipeline that pulls Geotab Trips API telemetry into Supabase (PostgreSQL) and normalizes raw JSON logs into query-ready tables using in-database SQL transformations, with tests covering data integrity across the ingestion path.
- Designed and built the PostgreSQL tables and REST endpoints behind the customer dashboards, including 7/30/90-day rolling fleet metrics, and generated the frontend client types from the OpenAPI spec so the UI and API stay in sync.
- Maintain the deployment setup: Terraform for Railway and Supabase environments, GitHub Actions for CI/CD and gated migrations, and ~42 scheduled workers keeping fleet data current across seven production clusters.
- Added a narration feature using the Anthropic API that turns fleet metrics into plain-language summaries for customers, gated behind a license check; built React/Tailwind dashboards to surface the metrics.
Elevate — Safety & Coaching Platform at MidnaTech (2025-04 – Present)
- Wrote the design docs, table design, and endpoint definitions for a multi-tenant platform that ingests vendor telemetry, groups events into incidents, and produces driver risk scores and coaching workflows.
- Designed the tenant isolation model — a separate PostgreSQL schema per customer with its own Alembic migration chain and row-level authorization — so one fleet's driver data cannot be read from another.
- Specified the reliability and privacy requirements the implementation is reviewed against: idempotent requests, retry limits, audit trails, and customer data export and deletion.
- Define endpoints as typed FastAPI response models, generate the frontend types from the committed OpenAPI spec, and keep a CI check that fails the build when the two drift apart.
Machine Learning & AI Intern at Kyndryl (2024-05 – 2025-01)
- Built a full-stack e-form parser (React, Node.js, Python Flask, PostgreSQL) that turns uploaded banking and financial forms into structured, searchable records, replacing manual re-keying and cutting form handling time by roughly 30%.
- Worked the full path end to end: extraction in Python, the Flask API in front of it, the PostgreSQL schema behind it, and the React interface reviewers used — so I could trace a field from the uploaded PDF through to the row it landed in.
- Designed the PostgreSQL tables and wrote the SQL to merge and compare records across submissions, so a new form could be checked against everything previously processed instead of reviewed on its own.
- Mapped extracted fields into sections mirroring the layout of the source form and surfaced them in AG Grid, letting reviewers confirm values against the original document instead of reading raw parser output.
- Set up CI/CD pipelines and version-control practices with Git and GitHub Actions for the team working on the system.
