Selected Work / AI Agents
ScentVerse
AI fragrance discovery companion using LLM, RAG, vector search and product recommendation logic.
AI AgentRAGVector DatabaseRecommendation UXFragrance Discovery
ScentVerse
Business Problem
- Online fragrance discovery is difficult because users cannot smell products.
- Traditional catalogues show notes and images but do not translate user mood, season or preference into product guidance.
- Users need a conversational assistant that can ask, interpret and recommend.
Solution
- Build an AI companion for fragrance discovery.
- Use LLM instructions, embeddings, vector retrieval and reranking.
- Connect poetic user language with product recommendation logic.
Workflow
- User intent, mood, season and preference are captured conversationally.
- Embeddings and vector retrieval connect natural language with fragrance/product data.
- Reranking and prompt instructions turn retrieval into explainable recommendations.
Tech / Tools
- Mistral
- OpenAI
- Pinecone
- Redis
- Python scraping / data engineering
- React TypeScript
- Vercel
- AI-assisted coding with Codex and Claude Code
Business Value
- Shows advanced AI/RAG product capability.
- Connects AI with commerce and product discovery.
- Proves domain curiosity in fragrance.
- Supports fragrance, beauty, AI product and no-code/low-code applications.
What It Proves
- Can build an AI product end to end, not only a prompt demo.
- Understands RAG as ingestion, embeddings, retrieval, reranking and UX.
- Connects AI with commerce and sensory product discovery.
Next Iteration
- Recommendation quality evaluation set.
- Feedback loop for liked/disliked suggestions.
- Product availability awareness.
- Fragrance memory profile.
- Admin panel for product data curation.
Related work
More proof from the same operating system.
Wholesale Operations Control Tower
Operations Systems
Live / 2026
Wholesale Operations Control Tower
Turns distributor, invoice and shipment data into prioritized blockers, next actions, follow-up drafts and weekly exception summaries.
B2B OperationsOrder-to-CashException Management
Catamalin
Data / Risk / Intelligence
Live / 2026
Catamalin
Aggregates French supermarket catalogues for consumers while creating a foundation for B2B retail promo intelligence.
Retail IntelligenceCatalogues PromoB2C Utility
French Company KYC Pre-Check
Data / Risk / Intelligence
Live / 2026
French Company KYC Pre-Check
Validates French company identifiers, normalizes company data, flags preliminary review points and generates structured reports for human review.
Fintech OpsKYB ReadinessPublic Data