I built Recorder Box as a heavy-duty recording and storage Device from scratch using linux ffmpeg stream recording package in a waterproof 12”x12” enclosure which comes in handy in Computer Vision projects during data collection phase usable in harsh conditioned environments like animal houses. The goal was not just to capture video, but to make footage easier to manage, monitor, retrieve, analyze and review when multiple DVRs and cameras are involved. A diversity of meshed cameras with different types – analogue or IP cameras – and standardized output video files for further computer vision are supported.
I pushed a new version of slidesMerge with several practical improvements based on real-world use. The core pipeline (coarse alignment, contour refinement, tiled warping, multi-channel merge) is unchanged. What changed is adding a centralized configuration file, improving edge-case handling in tissue detection, and improving resilience across diverse immunofluorescent slides with selectable image features.
I am publishing Patient Co-Pilot v2 Beta, a practical update focused on Multi-AI support and Symptom Monitoring. Download link for the Beta version is included.
I built Patient Co-Pilot mobile app to solve a common patient problem: confused with too much medical information, spread across too many places, from too many doctors.
I am pleased to share that our paper has been accepted at the 2026 International Conference on the AI Revolution: Research, Ethics, and Society (AIR-RES 2026) for both publication in the proceedings and conference presentation.
I developed ReadingOverlayTool to make long PDF reading sessions on macOS easier and less fatiguing. Super useful for putting less strain on eyes while reading long papers in any room lighting conditions.
I developed slidesMerge as a practical pipeline to align two large Aivia microscopy slides (>7 channels each) from different imaging cycles and fuse them into a single multi-channel OME-TIFF file.
I built this project to run a reproducible RNA-seq comparison between CD73 knockout (KO) and wild-type (WT) samples using a practical Kallisto + Sleuth workflow.
I built this project to improve background and artifact removal in multiplex pathology images, including whole-slide imaging (WSI) and multi-channel microscopy data.