Systems Software

Build a client-side auditor that tests post inclusion in LLM training sets with membership inference, n-gram lookup, and privacy-aware analysis.

Design a low-bandwidth lecture app, then test whether audio plus slide OCR improves comprehension and recall for remote learners.

Build and test a TinyML phone app that classifies gait signatures for navigation, and learn signal processing, model training, and field validation.

Build and test a personalized practice scheduler that models forgetting, compares learning curves, and measures long-term retention with A/B testing.

Design and test a scheduler that predicts task energy, compares CPU policies, and measures battery impact on mobile devices.

Build and test an auto-tuning key-value store, then measure how buffer size and workload mix change throughput on a Raspberry Pi.

Build and test cursor-based pseudo-haptic effects, then measure discrimination thresholds with simple psychophysics and clear data analysis.

Test how code watermarks survive renaming, reformatting, and other edits. Build evaluation skills in attack models, code analysis, and benchmarking.

Build and test column-aware Parquet compression ideas, then compare size, decode speed, and CPU cost on public datasets.

Build an online change-point detector for multivariate streams and test how well it finds regime shifts with low false alarms.