Projects, products, and useful implementation notes.
This page collects some of the things I've worked on and, where possible, the technical pieces behind them. If a library, architecture pattern, or small feature can be useful somewhere else, I try to call it out.
Manabi Kōbō
A free learning workshop for developers and other learners working toward practical Japanese reading and comprehension.
Useful pieces: OCR-assisted reading, Sudachi Japanese tokenization, furigana workflows, multilingual interfaces, guided courses, printable learning material, QR codes, PWA installation, local study tools, and AI-assisted content workflows.
Those components can also fit document processing, training systems, knowledge tools, and guided internal applications.
CodeSeal
A provenance registry for registering code-signing events and checking historical relationships while limiting exposure of sensitive signer information.
Useful pieces: Flask-based registry workflows, event history, source matching, notification tracing, controlled report access, and audit logging. Similar patterns apply to evidence trails, controlled historical lookup, and other provenance systems.
NAVI, monitoring & change utilities
A group of Python and SAP-oriented tools created around change analysis, job monitoring, reporting, Cloud ALM, RFC data access, and operational workflows.
Useful pieces: PyRFC/RFC integration, REST APIs, FastAPI/Flask services, Bokeh-style interactive reporting, dependency and impact logic, OAuth/XSUAA patterns, and lightweight operational interfaces around large enterprise systems.
Automation workbenches
Several generations of automation utilities cover SAP interaction, recorded desktop actions, data feeds, report creation, and newer browser-workflow ideas.
Useful pieces: action recording, parameterized steps, replay, diagnostics, file/data handoff, and keeping automation understandable enough to repair when the target system changes.
Specklet
A browser-owned planning and visualization workspace designed to break large goals, datasets, and projects into smaller pieces.
Specklet keeps its core experience deliberately lightweight: local-first project state, installable PWA behavior, structured task decomposition, reusable templates, responsive layouts, and simple visualizations. It is a useful reference for privacy-sensitive or offline-friendly personal tooling where a shared backend is not required.
NexoTori
A case-oriented information workspace that connects document capture, forms, review, automation, integrations, and operational insight.
NexoTori explores the full path from intake to action: document capture and OCR-style preparation, structured case data, forms, review queues, workflow automation, integrations, activity history, status monitoring, and dashboards. The Junbi Engine concept is the preparation layer that turns incoming information into something a team can validate and use.
ALM WebFlow Studio
A visual recorder and designer for turning browser interactions into reusable, inspectable automation flows.
WebFlow Studio combines live Chromium recording, captured objects, an object repository, data/variable handling, replay and training views, and execution history. Its hybrid boundary keeps browser control local while the studio provides the workflow model and review surface—useful for automation that needs to stay understandable rather than becoming a black-box macro.
ALM Engineering Hub
A standalone SAP COE engineering hub that brings change assurance, reliability tooling, reporting, guided workflows, and AI assistance into one extensible surface.
The Hub is built as a platform rather than a single dashboard: SAP BTP / Cloud Foundry deployment, XSUAA role control, destination and connectivity services, Python-backed utilities, Cloud ALM and Jira integrations, OData/RFC data access, installable PWA behavior, and Engenie as a routing/assistance layer. It is designed to complement specialist copilots and automation teams while giving ABAP and SAP engineers a practical path into Python and modern app patterns.
ProgreTech Mesh
A live, local control surface for observing and interacting with multiple autonomous agents without taking over their runtime.
Mesh is being developed around passive runtime observation, channel isolation, signed agent identity, heartbeat/status monitoring, agent-led enrollment, independent conversations, approvals, safe file exchange, and guided training. The design goal is plug-and-monitor: attaching Mesh should not restart, reset, or interrupt an agent already doing work on its own machine.
pAR3D
A real-time side-by-side 3D visualization experiment for desktop or camera feeds, aimed at consumer AR glasses and standard PC hardware.
pAR3D uses Python and Depth Anything v2 to estimate depth and synthesize SBS stereo output in real time. It supports desktop/camera input, adjustable depth and scale, performance modes, persistent settings, fullscreen/windowed operation, CPU/APU use, and optional CUDA acceleration. The project is a practical test bed for AI depth estimation, stereoscopic rendering, and wearable-display workflows.
Hardware, games & smaller experiments
Some projects exist because an old device, game, or idea was interesting enough to keep working.
This includes the Navia Dratp digital archive/remake, open tooling for legacy EinScan-S hardware, pAR3D stereoscopic experiments, and assorted small utilities. These tend to be good references for file formats, device integration, preservation, and exploratory Python work.