← All work

Marketplace · Next.js

Kova — personalized "Know Your Customer" marketplace

A D2C marketplace where customers define what they need — by chat, quiz, or filter — and a deterministic scoring engine returns every product with a 0–100% fit score and transparent "why this fits you" reasons.

0–100% match scoreExplainable AIMulti-vendorConversational UI
Industry
Multi-category Retail
Type
Personalized Marketplace
Stack
Next.js · NestJS · LangGraph
Scope
Discovery → Build → QA

The challenge

Products matched to the person, not the algorithm

Most recommendation engines are black boxes. Kova flips that: every match score is explainable, every reason is specific to the customer's stated needs. Customers engage via a structured quiz, a free-text chat, or traditional filters — all feeding the same scoring engine. Merchants get a vendor portal for catalog management and payout visibility.

What we delivered

  • Every recommendation score is explainable and cited
  • Multi-vendor marketplace with vendor self-service portal
  • Three input modes (quiz, chat, filters) → one scoring engine
  • Reduced returns through higher pre-purchase confidence
  • Live score refresh with optimistic UI
  • LangGraph-powered conversational discovery

Features

What we built

Six core capabilities that make the product what it is.

Deterministic Match Scoring

A rules-based engine that scores every product 0–100% against customer preferences. Every point is traceable — no black box, no guessing.

"Why this fits you" Reasons

Each result shows the top 3 reasons for its score: "Fragrance-free (matches: sensitive skin)" — building trust, not just ranking.

Multi-modal Input

Customers reach the same scoring engine via structured quiz, free-text chat (LLM-parsed), or traditional filter UI. All channels, one engine.

Multi-vendor Catalog

Merchants manage listings, variants, and pricing through a dedicated vendor portal with approval workflows and payout tracking.

Conversational Discovery

A LangGraph-powered chat interface interprets natural-language preferences and translates them into scoring parameters in real time.

Real-time Recommendation Refresh

Scores update live as customers refine their stated needs, with optimistic UI so the experience feels instant — no reload required.

Stack

Built with the right tools

No dogma about tools. We assessed what the product needed, then chose the stack that delivers fastest and stays maintainable long-term.

Next.js 14NestJSPostgreSQLPrismaLangGraphOpenAICeleryRedis

Ready when you are

Build something like this?

Tell us what you're working on. We'll map the right approach and stack for your context.