A selection of software projects that reflect my approach: technically rigorous, privacy-conscious, and grounded in a genuine theory of what the software should do for the human using it.
FamBot is a family coordination AI built on a privacy-first architecture — using locally-running Python code rather than cloud LLMs to analyze personal data. Rather than surrendering sensitive family communications to a remote service, the system processes everything on-device, preserving the intimacy of the data it handles.
At its core, FamBot builds a personalized family profile — extracting relationship signals, communication patterns, and coordination needs from existing data — and uses that profile to tailor its recommendations to each household's unique dynamics. The agent knows your family; it doesn't treat every user as a generic account.
The system reflects a deliberate philosophical commitment: that an AI handling the most personal domain of a person's life should be accountable, local, and genuinely oriented toward the family's flourishing rather than toward engagement or data extraction.
MusicSuaditor is a next-generation music recommendation system built around a philosophy of user flourishing rather than engagement. Where conventional recommenders optimize for time-on-platform, MusicSuaditor acts as a Suaditor — an advisory AI entity that genuinely serves your musical growth and satisfaction.
The system is built on BERT-style semantic embeddings, influencer knowledge graphs, and conversational onboarding that learns what music means to you — not just what you've clicked on.
"Suaditor" is a term I coined for a new category of AI advisor: one oriented toward human flourishing rather than algorithmic capture. The distinction matters. An engagement-maximizing recommender and a flourishing-maximizing recommender will, over time, recommend very different music — and produce very different people.
Past consulting and development work spans: Healthcare · Cybersecurity · IoT · Education