Selected Software Development Projects

Summary

Contributed to a client-based research software project extending OpenET, a Python framework for validating, visualizing, and benchmarking remote eye-tracking device data. The work focused on practical checks and plots that help researchers inspect data quality quickly.

What I built

• Designed Python data-quality checks for missing samples, duplicate timestamps, irregular sampling intervals, invalid gaze coordinates, incomplete recordings, and inconsistent metadata.
• Built visualization workflows for gaze traces, fixation stability, missing-data timelines, and sampling-frequency plots.
• Contributed user and developer documentation to make the tool easier for researchers to adopt.

Evidence

• Client-based project
• Client project: Dr Peter Wagner
• Data validation workflows

Tech stack

Summary

Designed and implemented a Dockerized format-aware black-box fuzzer for stdin-driven binaries. The tool mutates structured inputs and preserves reproducible crash, hang, slow-path, and unusual-output cases for later analysis.

What I built

• Built format-aware mutators for CSV, JSON, XML, JPEG, and plaintext inputs.
• Implemented timeout handling plus crash, hang, slow-path, and unusual-output capture for reproducible debugging.
• Packaged the workflow with Docker and Linux shell execution so targets could be tested repeatably.

Evidence

• Public GitHub repository
• Dockerized workflow
• Crash and hang capture

Tech stack

Summary

Built a Python simulator for a privacy-preserving contact tracing protocol with frontend nodes, a backend TCP server, and a passive attacker model. The simulator models rotating identifiers, encounter storage, upload/query flows, and adversarial observation.

What I built

• Implemented rotating ephemeral identifiers, UDP broadcast behavior, TCP request-response flows, and Bloom-filter encounter storage.
• Coordinated threading, locks, background loops, and interactive commands for node, server, and attacker workflows.
• Modelled passive attacker observation points to make privacy assumptions easier to inspect.

Evidence

• Public GitHub repository
• Protocol simulator
• Passive attacker model

Tech stack

Summary

Developed a team-based Android household inventory app with Firebase-backed authentication, storage, image handling, scanning, sorting, filtering, and UI testing. The app supports common inventory workflows from item creation through multi-selection and estimated-value tracking.

What I built

• Implemented inventory item flows using Android fragments, RecyclerView lists, ViewModel-style state management, and repository classes.
• Integrated Firebase Authentication, Firestore item persistence, and Firebase Storage image upload/delete workflows.
• Contributed UI tests for add/edit item, photo, scanning, sorting/filtering, multi-selection, and login behaviours.

Evidence

• Public GitHub repository
• Team Android app
• Firebase integration

Tech stack

Summary

Collaborated on a comparative computer vision benchmark for 12-class agricultural pest detection and classification. The project compared classical feature pipelines, hybrid approaches, and modern object detectors across accuracy and runtime metrics.

What I built

• Compared classical, hybrid, and deep learning approaches for 12-class insect detection and classification.
• Evaluated models using mAP@50, COCO mAP, accuracy, precision, recall, F1-score, ROC-AUC, latency, and FPS.
• Analyzed failure cases involving small, low-contrast, and camouflaged insects.

Evidence

• Report available upon request
• 80.77% mAP@50
• 97.73% classification accuracy

Tech stack

Summary

Conducted a digital forensic investigation of a simulated insider data exfiltration incident. The case study correlates disk, memory, registry, browser, email, USB, and network artefacts into a concise incident narrative.

What I built

• Correlated disk, memory, registry, browser, email, USB, and packet-capture artefacts into a single investigation timeline.
• Used hashing, artefact cross-checking, and timeline analysis to separate stronger indicators from system noise.
• Translated technical findings into remediation recommendations for account, endpoint, logging, DLP, and legal follow-up.

Evidence

• Report available upon request
• Reconstructed timeline
• Safe DFIR case study

Tech stack

Summary

Collaborated on a deep learning image-analysis pipeline for environmental microorganism images using EMDS-6. The project covered denoising, segmentation, feature extraction, classification, and object detection in a modular workflow.

What I built

• Compared denoising, segmentation, classification, and object detection methods in a modular image-analysis pipeline.
• Evaluated segmentation and classification with task-appropriate metrics including IoU, Dice, and F1-score.
• Connected preprocessing quality to downstream model performance through comparative experiments.

Evidence

• Report available upon request
• 0.9365 IoU
• 0.9843 Dice

Tech stack