A platform for annual planning and performance tracking, used primarily by value analysis and engineering teams. Goals are defined at executive level and cascade through senior managers, managers and team leads to individual contributors, who submit plans and actuals each period. Submissions are aggregated and rolled up by team and division, with each goal type — financial, rating and specific — following its own calculation and roll-up rules. As technical owner I was responsible for the application from requirements through to release, including system architecture, REST API design, database design with application-driven migration, and the React front end. Implemented modules cover user management and organizational relationships, role-based access control, goal and metric management, goal cascading, timeline management for deadlines and year closing, and the plan and actual submission and approval workflow. Supporting work included JWT authentication, scheduled background jobs for reminders and periodic calculation, logging, email notification, and data export to Excel and PDF. The application is in use by the client, with a second phase planned.

React · C# · ASP.NET Core · MySQL · JWT · Quartz

A JavaFX-based Windows HMI for a chromatographic system stored its data through an XSD-based model, and the libraries supporting that approach were being withdrawn. The objective was to move persistence to PostgreSQL and MySQL and place the product on a more durable footing. The architecture was proposed by the manager and approved by the client. I reviewed the existing model — 150 to 200 types, with several levels of inheritance, nested objects and lists of nested objects — documented it through ER diagrams and mind maps, and validated the analysis against sample models before implementation began. I then developed the serializer and deserializer that persist the live object model to the database and reconstruct it on load, and implemented an equivalent library in C++ for a second product variant so that both share a single persistence contract. I was primary developer on both implementations, with one colleague contributing to analysis, mapping and implementation. Both were delivered and integrated into their respective products, and the work was recognized with a Pat on the Back award for the quarter.

Java · JavaFX · C++ · PostgreSQL · MySQL

A proof of concept demonstrating monitoring and control of a connected appliance from a mobile device. The application pairs with a refrigerator by scanning a QR code, then reports refrigerator and freezer temperature, humidity, door and ice maker state, power consumption and historical trends, along with warnings and errors. Control functions include eco, fast and defrost modes, time scheduling, and management of several appliances from a single application. I implemented both ends: the Jetpack Compose application and a Python simulator standing in for the appliance, with MQTT as the transport between them. The work was presented to the client as a demonstration of capability.

Android · Jetpack Compose · MQTT · Python

A vehicle HMI dashboard produced in two variants: an Android touch display with map navigation, and a smaller non-touch ESP32 cluster. I implemented the Android dashboard front end, including a custom canvas-based speedometer, drive modes, odometer, throttle and tell-tale indicators, with data and logic prepared at the ViewModel layer so that only the values required for display reached the UI. I subsequently took responsibility for the CAN parsing layer: interpreting the client-supplied signal specification, handling endianness, scaling factors and derived calculations, and mapping roughly ten signals arriving at intervals between 1 and 50 milliseconds onto the vehicle data model. Output was checked first against expected values worked out in a spreadsheet, then against recorded CAN logs supplied by the client. I also wrote a Python utility for Windows that generates CAN frames from the specification and transmits them onto the bus, so that the parsing layer and the dashboard could be exercised without a real vehicle. Parsing and presentation ran as separate processes communicating over AIDL, so that either could restart independently. Local storage and periodic MQTT publication were added for client-specific requirements.

Android · Java · Kotlin · Python · CAN · AIDL/IPC · SQLite · MQTT

Recognition of the dashboard and CAN work led to a move to the research and development team, where I replaced the SDK responsible for communication between the companion application, the Android dashboard and the ESP32 cluster. The existing implementation was unreliable, had not kept pace with newer Android releases, and could not accommodate the range of requirements then in scope. I analyzed the existing architecture, investigated the reported defects, and designed and implemented the replacement under the guidance of the team lead. The SDK covers pairing, connection management, request and response with acknowledgement, retry, and a queue that permits a single exchange in flight at a time — a constraint set by the ESP32, which had approximately one kilobyte of usable heap available. It also supports firmware update over the air.

Kotlin · BLE · ESP32 · C · C++

Parses an XSD schema and generates the model classes it describes, including their types and default values, together with the shared serializer and deserializer functions, removing the need to write them by hand.

Generates CAN frames on Windows from the signal specification and transmits them onto the CAN bus, so that the parsing layer and the Android dashboard can be exercised without a real vehicle.