Tom – Golang, Kubernetes, Node.js

The project was a complex internal Apple project to consolidate two authentication and authorization systems by migrating the legacy IAS (Golang-based) platform to IdMS. Both systems provided AuthN, AuthZ, and IAM services, but with very different implementations. The migration involved ensuring a transparent transition for thousands of internal users, handling challenges such as mTLS certificate management, DNS record updates, and OIDC/OAuth2 endpoint swaps, all while keeping the legacy system operational until full decommissioning in FY25.

Tom's main scope of work included:

• led the migration of the legacy IAS (Golang-based) authentication/authorization system to IdMS, ensuring a seamless transition for thousands of internal users;
• architected and implemented facades, proxies, and compatibility layers to address key differences in AuthN/AuthZ flows, particularly around claim issuance in identity tokens;
• ensured zero downtime during client migration by carefully managing cutovers, OIDC/OAuth2 endpoint swaps, mTLS certificate renewals, and DNS updates;
• unofficially directed a small team (4 engineers), effectively taking on roles of principal developer, team lead, and project coordinator;
• improved test coverage from ~30–40% to 90%, significantly increasing system reliability and maintainability;
• maintained legacy IAS system in parallel, resolving semi-critical bugs and ensuring continuous service until full decommissioning;
• coordinated with stakeholders to align on migration milestones, risk management, and FY25 project completion targets;
• managed ongoing IAM data migration, ensuring smooth progress with minimal disruption.

A peer-to-peer, real-money sports gaming platform tailored for friends and groups across the U.S. It offered skill-based contests—such as Survivor, PickX, and Tiers—where participants could compete with entry fees and cash prizes. Built on the legacy of RunYourPool and Office Football Pool, the platform modernized traditional sports pools with integrated payment handling, leaderboard management, and compliance across multiple states. It featured a "Commissioner Economy™" model to reward contest organizers, and quickly expanded to over 33 states with millions of users. All iterations were handled as a standalone product by Splash Inc.

Among others, Tom contributed in the following ways:

• drove the project from zero to a working proof of concept within one month;
• redesigned the CI/CD pipeline to significantly improve efficiency and reliability;
• established guidelines for testing, documentation, and development best practices;
• built the critical "Scores" service, aggregating real-time feeds from multiple sports data providers to power leaderboards, odds calculations, and live updates;
• engineered the service as a polyglot system (Nest.js/Node.js for microservice integration, Golang for federated GraphQL queries, and exploration of Rust for performance optimization);
• ensured compliance with state and federal gaming regulations, integrating auditing processes into the OTEL telemetry stack;
• enabled continuous monitoring for auditing, security, and performance issues, surfacing insights for the SRE team;
• acted as a point of escalation for the SRE team post-launch to resolve critical issues quickly.

Peacock is an American over-the-top (OTT) streaming service owned by NBCUniversal, offering a diverse blend of television shows, movies, live sports, news, and original programming. Launched in July 2020, it featured content from the NBCUniversal library—such as classic sitcoms, Universal Pictures films, and premium originals—alongside live sports like NFL, Premier League, Olympics, and WWE.

Tom's contributions included, but were not limited to:

• revamped the BRB project from a brittle HTML/JavaScript proof-of-concept into a modern, production-grade platform using React, GraphQL, and supporting backend services;
• built SCTESpy, a Golang service to capture and decode MPEG-TS frames (via UDP multicast) and detect SCTE-35 ad avail markers for real-time ad insertion;
• reverse-engineered the REST protocol for hardware audio/video encoders, requiring deep research into SCTE specifications to enable reliable integration and control;
• designed and implemented a Nest.js backend service for persistent monitoring of encoder operating characteristics, feeding data both to the frontend and NBCUniversal’s internal telemetry/observability systems;
• developed a React-based real-time dashboard to provide studio producers with a modern equivalent of the “big red button,” integrating live telemetry, encoder controls, and SCTE event triggers;
• enabled real-time ad avail insertion, alternate feed switching, and AV processing (e.g., audio downsampling) through coordinated communication between frontend, backend, SCTESpy, and hardware systems;
• modernized and stabilized an early-stage tool into a production-grade system still in active use for Peacock’s live streaming ad operations.

The project was an AI-powered chatbot during the COVID-19 pandemic. Initially focused on epidemic response, it used natural language processing (via Azure NLP) to guide users through physician-designed symptom triage flowcharts. Over time, the platform evolved into a general-purpose patient assistant for scheduling appointments, checking lab results, refilling prescriptions, and exploring care options.

The scope of work was:

• created overall render loop for Grace, moving from raw DOM manipulation into React-based flow;
• built the backend for the Covid symptom flow, as well as rationalized the entire backend structure using the Azure NLP functionality;
• unblocked the designers who were previously heavily constrained by the raw DOM manipulation in terms of how they could display content;
• added some Golang microservices in performance-critical areas (e.g. 911-handoff, etc.);
• made the entire system deterministic and testable.