Notemae

The description states that Notemae is a bilingual (English and Indonesian) web app with four connected experiences:

1. A fragrance catalog you can browse and filter.
2. A dupe guide built around three steps: pick the original perfume, read a similarity score computed from shared catalog notes rather than ad claims, then compare candidates side by side.
3. An AI consultant that accepts requests in everyday language and turns them into typed preference profiles, returning recommendations over server-sent events.
4. A system designed to never invent products, only reorder and explain existing entries in a PostgreSQL catalog.

It is described as an editorial experience, using technologies like React 19 + TypeScript, Express SSR, Go backend, Python FastAPI agent, Qwen LLM via DashScope, pgvector for vector retrieval, and Docker Compose for deployment.

Inference The system appears to be a hybrid of deterministic logic and LLM-based reranking, where constraints are enforced before model use, and explanations come only from existing catalog data.

The description states that Notemae aims to be a "guide that listens like a friend who is deep into fragrance, but only ever answers from a real catalog it can show you."

It positions itself as:

• A personal scent consultant.
• An editorial, bilingual web app.
• A dupe guide that uses similarity scores based on shared notes rather than marketing claims.
• A grounded AI assistant, never inventing products or performance claims.

The authors also note they were inspired by the difficulty of buying fragrances online and wanted to bridge the gap between human language and fragrance catalog data.

Inference The product evolved from a simple idea (bridging language gaps) into a full-stack system with AI, catalog, and user interface components. It reflects an attempt to create a trustworthy, data-driven recommendation tool.

The description does not clearly define the target customer or Ideal Customer Profile (ICP).

It implies that users are people looking for fragrance recommendations based on:

• Personality traits
• Climate conditions
• Occasions
• Budget constraints

However, there is no explicit segmentation or persona definition beyond general descriptions like “someone who wants to buy a fragrance online.”

Inference Users likely fall into a niche market of fragrance enthusiasts or casual buyers seeking personalized advice.

There is no evidence provided regarding the business model or pricing strategy.

The description does not mention:

• Revenue streams
• Monetization plans
• Subscription tiers
• Paid features
• Licensing models
• Retail partnerships

Inference Given the hackathon context and lack of traction evidence, it is unclear whether Notemae intends to operate as a freemium service, a marketplace, or a B2B tool for retailers.

The project includes detailed technical architecture:

• Frontend: React 19 + TypeScript SPA served by Express with SSR.
• Backend: Go-based clean architecture using pgx and standard library HTTP stack.
• Recommendation Agent: Python + FastAPI service with Qwen reranking, Borda count, and diversity control.
• Retrieval Engine: PostgreSQL 16 with pgvector for vector search.
• Ingestion Pipeline: Scheduled worker pulling from approved sources (e.g., Open Beauty Facts), normalized and enriched by Qwen.
• Deployment: Docker Compose setup.

Key engineering decisions:

• Rewrote backend from FastAPI to Go, ensuring byte-for-byte parity.
• Used deterministic constraint enforcement before LLM calls.
• Built evaluation harness for stochastic reranker.
• Implemented SSR without breaking JSX transformation.

Inference The system shows strong engineering maturity for a hackathon project. It demonstrates careful attention to correctness, scalability, and maintainability.

There is no evidence of traction or user adoption beyond the hackathon submission.

The description states:

• This was submitted to the OpenAI 2026 hackathon.
• No revenue, customer, or usage data is available.
• The authors mention offline A/B testing results but do not present production metrics.

Inference This is a prototype or proof-of-concept built for a competition. There is no evidence of product-market fit or commercial viability at this stage.

The description does not provide any information about competitors or market positioning relative to others in the fragrance or personal care space.

It does not reference:

• Existing fragrance recommendation platforms
• Similar AI-powered shopping tools
• Retailer or marketplace integrations
• Industry benchmarks or pricing models

Inference Notemae appears to be a new entrant in the niche of personalized fragrance guidance, potentially targeting users who struggle with online fragrance selection. Its uniqueness lies in its grounded AI approach and bilingual support.

Several risks and red flags are evident from the self-reported description:

1. No traction or revenue: The project is described as a hackathon submission with no evidence of real-world usage.
2. Limited scope: Only one team member (2 people) built it, suggesting limited scalability or long-term commitment.
3. Unproven commercial viability: No business model or monetization strategy discussed.
4. High technical complexity without validation: While the architecture is impressive, there’s no evidence that it works in production at scale.
5. Dependency on external data sources: Reliance on Open Beauty Facts and licensed feeds may limit catalog completeness or introduce quality issues.
6. Lack of product-market fit signal: No indication that users actually engage with the system beyond testing.

1. What is the current status of the product? Is it live or still in prototype phase?
2. How many users are currently using Notemae, and what is their engagement level?
3. Have you conducted any user research or usability testing beyond the hackathon?
4. What are your plans for monetization and scaling the business?
5. Are there any legal or licensing issues with the data sources used (e.g., Open Beauty Facts)?
6. How do you plan to expand the fragrance catalog, especially for underrepresented regions or niche categories?
7. What is the long-term roadmap beyond the current features?
8. Can you share more about how the AI model was trained and validated?