Experience

Roles and education. Download the full CV as a PDF.

Work

Software Engineer

  • Built a .NET 10 Android binding of Garmin's native Health SDK (Standard SDK 4.7.0) and a .NET MAUI companion app that pairs Garmin wearables and streams 7 live vital-sign types (heart rate, HRV, SpO₂, respiration, stress, steps, accelerometer) to the VitalsIQ backend.
  • Authored the Java-to-C# binding transforms resolving R8, Guava, SQLCipher, and Gson interop conflicts, plus an Android foreground service for resilient background sync.
  • Built a CivTAK (civilian ATAK) plugin in Java that surfaces wearable vitals as an in-app pane and map marker, on a device-independent, JUnit-tested core with a bound-service Messenger IPC bridge to the MAUI app.
  • Ported the full JABCode colour 2D-barcode library (encoder and decoder with error correction) from C to C#.

Embedded Software Engineer

  • Built the end-to-end noise-classifier model pipeline (data preparation, training, evaluation, and firmware integration) for on-device inference on constrained hardware; the classifier achieved 92% accuracy at a size of only 40 KB.
  • Developed AuraSense firmware in C/C++ on Zephyr RTOS / nRF Connect SDK for a solar-powered nRF54L15 + nRF9151 noise and air-quality IoT monitor, architecting 11 self-contained driver components (~3K LOC).
  • Built a real-time acoustic sound-level-metering pipeline: A-weighted (3-stage biquad IIR) dBA and Leq over 10 s windows from stereo 16 kHz PDM MEMS microphones, with calibrated dB-SPL and clipping detection.
  • Cut measured idle current from ~960 µA to ~189 µA via PPK2 differential power profiling, root-causing the LTE-modem floor, an LDO-vs-DC/DC path, and a mis-wired PMIC GPIO.
  • Implemented an LTE-M / NB-IoT uplink (nRF9151 serial modem, AT sockets, hand-written JSON codec) to a Node.js / Express telemetry backend.

Machine Learning Engineer Intern

  • Built an AI document-ingestion pipeline to parse PDFs, spreadsheets, and images (with OCR), then auto-summarise and classify content, cutting manual screening from 10 minutes per document to 2.
  • Shipped a REST service and lightweight web UI that streamlined analyst review, cutting document navigation and handling time by roughly 60%.

Machine Learning Engineer Intern

  • Designed an English keyword-spotting pipeline; precision went from 92% to 94% and WER dropped by 1.1%.
  • Built, cleaned, and validated large-scale audio datasets with forced alignment and transfer learning; expanded coverage from 100 to 300+ hours and reduced corrupted data from 10% to 2%.
  • Built a reproducible benchmarking pipeline for ASR/TTS candidates, cutting evaluation time from 4 hours to 1.

Education

MSc Data Science & Artificial Intelligence Technology

BSc (Honours) Computer Science & Engineering

  • Honours Programme. Minor in Finance (Sep 2025 – Jan 2026).