Python, Django, React, Next.js, TypeScript, Snowflake, PostgreSQL, Redis, Docker, AWS Lambda/S3, GitLab CI/CD, Datadog

• Principal engineer on the linear and digital audience forecasting platforms that project episode-level viewership across Disney's media networks, which drives advertising inventory decisions. Authored roughly 2,100 of 8,300 commits on the core service.
• Drove AI-assisted development: authored the agent instruction and skill files that ship with the project template, so every repository inherits them, and built MCP servers exposing internal tooling to coding agents. Worked daily using Claude Opus.
• Led the migration of the forecasting data layer from PostgreSQL to Snowflake: key-pair authentication, an identifier-casing migration behind a backwards-compatible view layer for existing consumers, and a move off a forked database driver onto the official upstream one.
• Designed and shipped roughly a dozen reusable internal Python packages on a private index: OIDC/SSO auth, AWS Parameter Store configuration, middleware, health checks, Snowflake schema models, and PEP 561 stubs. All fully typed, and they drove an audit that pushed infrastructure code out of eight applications into shared libraries.
• Built a package that deploys containerised Django and Next.js applications to AWS Lambda, turning a bespoke per-project deployment into a dependency plus a CI job.
• Rebuilt the test strategy across services: replaced database-dependent tests with zero-database integration tests, raised backend coverage from 51% to above 90% with enforced thresholds, and reached 100% on several services. Migrated the frontend from Create React App to Vite and Vitest, added the first frontend test suite at 92% line coverage, and moved it to TypeScript strict mode.
• Owned CI/CD and tooling standards: consolidated projects onto shared GitLab CI templates, wrote the code-quality report generators that surface Python formatter and linter output natively in merge requests, and created the project template system that keeps every repository current on dependencies, linting, typing, and testing configuration from a single update.
• Integrated Datadog APM tracing and RUM across the web applications, migrated caching to managed Redis (ElastiCache cluster mode), and served as the go-to person for deployment, CI/CD, and Python debugging across teams, mentoring junior developers.

TypeScript, JavaScript, Node, React, PostgreSQL, AWS S3/RDS/CloudFront, GitLab CI

• Owned the company's Zapier integration end to end (over 540 commits), exposing telephony, messaging, and meeting APIs to non-developers. Converted polling triggers such as call-ended and new-voicemail into webhook-driven instant triggers, cutting latency and API load.
• Built the integration's TypeScript functional-testing framework and the CI/CD around it: suites triggered on private builds, coverage reported to GitLab Pages and InfluxDB, and automated release promotion. Instrumented the product with Mixpanel and Segment analytics, and fixed OAuth token lifetimes that were silently expiring connections.
• Contributed to a new Microsoft Teams integration (automatic app installation into Teams, user account linking) and to a React presence-synchronisation frontend. Maintained the Zendesk integration for making and tracking calls from support tickets.

TypeScript, React, Next.js, Flask, PostgreSQL, PyTorch, CUDA, Docker, nginx, uWSGI, AWS CDK/Lambda

• Sole primary engineer on a full-stack marketing campaign-measurement application (372 of 392 commits): React and TypeScript frontend, Flask REST API, PostgreSQL schema, and the analyst-facing interface for three causal-inference report types covering synthetic-control comparisons, covariate balance, and per-region weights.
• Built the GPU inference service behind it (Flask, uWSGI, PyTorch, HuggingFace Transformers on an NVIDIA DGX system): one model pipeline per detected CUDA device, requests routed to the first free device, HTTP 503 when all were saturated. Diagnosed CUDA failing to initialise under uWSGI's pre-fork worker model and fixed it by deferring model loading into a post-fork hook.
• Designed that application's report-request pipeline: JSON-Schema-validated intake, a single transaction writing the request and its parameters, dispatch to an external modelling service, and rollback on either a database error or a failed dispatch, so a failed run never left orphaned rows behind.
• Shipped features to a Next.js news-analytics product on AWS Lambda@Edge, and wrote a standalone AWS CDK application that codified a hand-built account as infrastructure as code (VPC, public and private subnets across four availability zones, NAT gateways, route tables, DynamoDB), correcting a routing defect in which the NAT gateways sat on private subnets.

• Delivered the ground-up overhaul of the same forecasting platform, over 1,500 commits. Wrote nearly all of the test suite, taking coverage from a low baseline to above 95% across frontend and backend, and completed the backend type annotations to make static analysis viable.
• Architected the reporting application, added the GitLab CI hooks for QA and coverage reporting, and managed the nginx configuration used to host the site.

This contract ended because of cutbacks due to the Coronavirus pandemic. I was invited back in 2023.

Python, asyncio, Oracle, AWS Lambda, mypy, pytest

• Built an asyncio Python service synchronising survey panelist records between an Oracle data warehouse and a third-party sample-exchange API, deployed to AWS Lambda, fully type-annotated and tested.