CigarCompanion
Personalization in Subjective Domains
CigarCompanion is an AI-powered personalization system designed for a subjective, taste-driven domain where trust, frictionless UX, and data ambiguity are core constraints.
Designed a canonical data model for cigars in a domain with no reliable standard and conflicting sources
Built multi-source ingestion, normalization, and entity resolution pipelines to establish a durable catalog
Treated AI as an interface reducer, using voice, vision, and natural language to eliminate manual interaction
Designed uncertainty-aware workflows (confidence scores, suggestions, human validation, controlled entity creation)
Implemented privacy-first AI surfaces, with clear boundaries between on-device and cloud intelligence
Built personalized recommendation logic for a domain where "best" depends on user intent, culture, and context
Designed for progressive trust and adoption, favoring a large freemium surface before monetization
Developed both a personal companion app and a public knowledge platform on a shared canonical core
Montecristo No. 2
"Rich wood and spice notes with a creamy finish. A masterpiece of construction."
"CigarCompanion demonstrates my ability to design AI systems for subjective domains — where personalization, trust, and human preference matter more than optimization."