GlamBought

Inspiration

Beauty shopping is overwhelming—too many products, too much noise and constant novelty. Whether you’re Jarmar, a skincare newbie, or Angela, a global beauty veteran, finding what actually works for you is hard. The right product depends on your skin, hair, climate, budget, past experiences, travel, lifestyle, and what you already own — yet most shopping experiences make you start from scratch every time.

We built GlamBought around a simple question:

What if you had a personal beauty shopper that actually knew you?

An agent that remembers what works for you, anticipates what you might need, researches the live market, and turns hundreds of options into a decision you can actually act on.

We didn't build an AI that knows beauty. We built an AI that knows your beauty.

Beauty, done for you.

What it does

GlamBought is your personal AI beauty shopper.

Instead of navigating endless filters, reviews, Reddit threads, TikToks, ingredient lists, and retailer pages, you simply tell GlamBought what you need.

GlamBought combines your personal context — skin and hair profile, preferences, purchase history, budget, location, travel, calendar, and connected lifestyle data — with live beauty intelligence.

It researches products, ingredients, reviews, prices, promotions, and availability; evaluates which evidence is trustworthy; and recommends the products that best fit you and your current situation.

It can also be proactive.

If your calendar shows you're hiking in Switzerland next week, for example, GlamBought can anticipate that the colder, drier climate, higher elevation, outdoor activity, and UV exposure may change what you want to pack. It can then research appropriate products and build a personalized routine within your budget — before you even remember to ask.

How we built it

We designed GlamBought around four core layers:

1. Cognee — Personal Brain

Cognee gives GlamBought memory. It stores structured context about the user: beauty profile, preferences, routines, product experiences, purchases, and relevant lifestyle context.

2. Bright Data — Live Beauty Intelligence

Bright Data connects GlamBought to the live web so it can research current products, ingredients, retailers, prices, promotions, reviews, beauty communities, and creator content instead of relying solely on static model knowledge.

3. AWS Strands Agents — Reasoning & Orchestration

Strands powers the agentic workflow. It interprets a natural-language request, determines what information it needs, retrieves relevant personal context, calls research tools, evaluates candidates, and turns the results into an actionable recommendation.

4. Docker Sandbox — Deterministic Computation

The sandbox handles deterministic tasks such as normalizing product and pricing data, comparing candidates, applying constraints, and optimizing a basket against a user's budget.

Together, they create a continuous loop:

Memory → Context → Research → Reasoning → Recommendation → Action → Better Memory

Challenges we ran into

Our biggest challenge was scope.

Beauty is enormous: skincare, makeup, haircare, thousands of products, global trends, changing prices, conflicting reviews, and highly personal preferences.

We had to resist building a giant beauty search engine and instead focus on the agentic experience:

Can GlamBought understand me, do the research I don't want to do, and reduce it to a useful decision?

Another challenge was trusting beauty data.

A dermatologist, ingredient database, long-term Reddit review, verified purchase, sponsored TikTok, and retailer description shouldn't all be treated as equally reliable. We needed to think carefully about source provenance, commercial influence, expertise, and confidence.

Live commerce data is messy too. The same product can have different names across retailers, prices and promotions constantly change, and availability depends on location.

Finally, we had to think carefully about personalization versus privacy. Calendar, travel, purchase, and wellness context can make an agent dramatically more useful, but sensitive information should only be used with permission and for a clear purpose.

Accomplishments that we're proud of

We're proud that we turned a broad idea — "AI for beauty shopping" — into an agent that can reason across personal memory, real-world context, live beauty data, and shopping constraints.

Rather than building another chatbot that recommends popular products, GlamBought can understand two completely different people and behave differently for each.

For someone who knows almost nothing about skincare and wants maximum simplicity, GlamBought can cut through thousands of products and build a three-product routine.

For an experienced beauty shopper traveling internationally, it can anticipate an upcoming trip, account for climate and activities, understand detailed ingredient and skincare preferences, respect category-specific price sensitivity, research the live market, and build a personalized shopping plan.

Most importantly, we built the foundations for an agent that doesn't just answer:

"What should I buy?"

It learns:

Who are you? What has worked for you? What's happening in your life? And what are you likely to need next?

What we learned

Our biggest lesson was that personalization isn't just adding someone's skin type to a search query.

A useful personal agent needs to understand context.

The exact same person may need completely different products in humid Taiwan, dry San Francisco, or while hiking at elevation in Switzerland.

Likewise, a product can be objectively excellent and still be the wrong recommendation because it's too expensive, unavailable nearby, contains something the user avoids, duplicates something they already own, or simply doesn't fit their lifestyle.

We also learned that beauty intelligence is global. A North American shopper may discover her best product through Korean beauty, Japanese beauty, European skincare, a dermatologist, Reddit, or a small creator halfway around the world.

The agent's job isn't just to find information.

It's to determine which information matters for this person, right now.

What's next for GlamBought

Our vision is for GlamBought to become a beauty agent that gets better every time you use it.

It should remember what you bought, learn what actually worked, understand what you repurchase, know when you're likely to run out, watch for better prices, anticipate upcoming needs, and prepare purchases for your approval.

Over time, GlamBought could move from:

"Find me a moisturizer."

to:

"Your moisturizer is probably running low. It's 25% off today at your preferred retailer, and you have a trip next week. Want me to add it to your order?"

The goal isn't to give people more beauty content.

It's to eliminate the work between:

"I need something." → "I got exactly what works for me."

GlamBought — Beauty, done for you.

Built With

• amazon-web-services
• brightdata
• chatgpt
• claudecode
• cognee
• github