Joakim Kanichai

• Assisted in maintaining cleanliness of dining areas and food service stations.
• Operated kitchen equipment to efficiently prepare food items.
• Collaborated with team members to ensure smooth dining operations.

Timing-Jitter Robust Control Research (Python, PyTorch).

• Built a Python simulation and benchmarking framework to evaluate controllers under variable-rate timing jitter across multiple jitter levels and disturbance conditions, with multi-seed (50) experiments.
• Implemented and compared learning-based residual controllers (MLP, LSTM, continuous-time/LTC-style) against a PID baseline; tracked failure rate, RMSE, and steady-state variability.
• Developed a reproducible evaluation pipeline: automated sweeps, CSV logging, and figure generation for results.
• Performed statistically rigorous analysis using bootstrap confidence intervals and Mann–Whitney U tests, with Holm correction for multiple comparisons.
• Produced paper-ready visualizations and structured research documentation (methods, protocol, results).

Semantic Drone Navigation Using CNNs (ME459)

• Developed a CNN-based perception pipeline for navigation using a pretrained ResNet-18 backbone (transfer learning).
• Reduced overfitting by freezing the backbone and training the final layer on labeled image crops.
• Trained with cross-entropy loss using the Adam optimizer, and applied data augmentation for robustness.
• Demonstrated a navigation simulation that uses visual classification to support correct path/target identification.