Personal outfit & style recommendations

Outfits that already
know you.

Hemline reads your style profile, your real wardrobe, the occasion and today's weather, then suggests complete looks and explains why each one works. Every like and skip makes the next suggestion better.

0pieces in wardrobe
0outfits evaluated per request
0looks kept in beta

How it works

Four steps from closet to confidence

01

Profile

A two minute quiz captures style direction, palette, fit, budget and the occasions you dress for.

02

Wardrobe

Pieces are tagged by category, color, formality, warmth and style, with auto tagging from a photo.

03

Recommend

A hybrid engine scores thousands of combinations against occasion, weather, color harmony and taste.

04

Learn

Likes, skips and what you actually wear re-weight the model so it keeps getting closer to you.

Step 1 / Style profile

Tell us how you like to dress

Step 2 / Recommendations

Today's outfits

The engine is learning

Like or skip a look and watch the weights move. Liked pieces and color pairs get a boost, skipped ones fade, and items worn in the last 3 days are rotated out.

No feedback yet. Try the like and skip buttons on a look.

Step 3 / Digital wardrobe

Your wardrobe, tagged and searchable

Step 4 / Week planner

A full week, no repeats

Fill the gaps

Buy less, wear more

The engine measures how many new outfits each missing piece would unlock, inside your budget, so every purchase earns its place.

Under the hood

A hybrid, explainable recommender

Inputsprofile, wardrobe, occasion, weather API, calendar
Candidate generationrule filters + CLIP image embeddings (pgvector)
Scoringoccasion, weather, color harmony, taste, novelty
Re-rankdiversity (MMR) + collaborative signals
Feedbacklikes, skips, worn logs update weights

Model metrics offline eval, last 30 days

Recommendation API

// POST /v1/recommendations
{
  "user_id": "u_2841",
  "occasion": "client_meeting",
  "weather": { "temp_c": 9, "rain": true },
  "limit": 3
}
// 200 OK  (p95 84 ms)
{
  "outfits": [{
    "items": ["shirt_white", "chino_navy", "blazer_navy", "derby_brown"],
    "score": 0.91,
    "why": ["formality match", "tonal navy palette"]
  }]
}
Next.js + React NativeFastAPI (Python)PostgreSQL + pgvectorCLIP image taggingLightGBM rankerRedis cacheWeather APIDocker / AWS

Delivery plan

From MVP to a learning product

  1. Week 1

    Data model & profile

    Item schema, style quiz, auth, wardrobe CRUD with photo upload.

  2. Week 2

    Engine v1

    Rule based scoring for occasion, weather and color harmony, explainable output.

  3. Week 3

    Feedback & planner

    Like / skip loop, worn log, weekly planner, gap analysis.

  4. Week 4

    ML & launch

    Image auto tagging, learned ranker, metrics dashboard, deploy and handover.