Vintage commuter96% match
Search infrastructure for resale
Make every
one-of-one item findable.
Second-hand search is not catalog search. Advec turns imperfect photos, inconsistent listings, and constantly changing inventory into relevant discovery experiences.
Built from hands-on work in fashion, visual media, and high-volume content platforms.
Search by image
Find items like this
Relevant experience with
Artlist
Lightricks
YAGA
BrandAlley
Digital Turbine
The marketplace reality
A living inventory needs a different search engine.
General-purpose search is usually tuned for stable, standardized catalogs. Resale marketplaces are the opposite: every listing is unique, incomplete, visual, and temporary.
Amateur content in, usable signals out.
Sellers take quick phone photos under inconsistent lighting and write short, noisy descriptions. We extract the signals buyers actually need.
No standard catalog
AI categorization and attribute extraction turn free-form listings into a taxonomy you can search, filter, and measure.
One item, then gone
Most items have a quantity of one. Ranking and availability must react the moment something is listed, reserved, or sold.
Inventory that breathes
High-frequency updates are a core product constraint, not a background sync problem. We design retrieval around the churn.
The photo is often the best metadata
Native image search and multimodal ranking help buyers find what they mean, even when sellers describe it differently.
What we build
From camera roll to conversion.
One connected discovery stack, designed around the reality of user-generated inventory.
Capture
A seller uploads a photo and a few words.
Enrich
Models infer category, attributes, style, and visual vectors.
Retrieve
Hybrid text and image search surfaces the right candidates.
Rank & learn
Session signals improve relevance while inventory keeps moving.
Image search that ships as a product feature.
Find visually similar listings, deduplicate content, and let buyers search from a photo — with the indexing and ranking layer included.
Structure without burdening the seller.
Automatically classify categories and extract attributes from text and images, while keeping your taxonomy adaptable to real inventory.
Ranking for items that may disappear today.
Blend semantic, lexical, behavioral, and availability signals in a retrieval system that reacts to new and sold inventory quickly.
Personalization before a long history exists.
Use session-based and item-to-item models to create relevant feeds, alternatives, and “more like this” journeys from sparse signals.
Relevant experience
The hard parts are familiar.
Our work spans visual retrieval, fashion discovery, recommendations, and high-volume categorization — the same building blocks second-hand marketplaces depend on.
01Fashion search
YAGA
Search for a fashion resale marketplace.
Marketplace-specific discovery for unique fashion listings and the inconsistent seller data that comes with them.
Visit marketplace02Visual search
Artlist
Visual reranking for creative inventory.
Matrix factorization plus visually driven reranking made recommendation feeds more cohesive while preserving artistic standards.
Longer sessionswhile maintaining relevance
03Video recommendation
Lightricks
Multimodal recommendation at scale.
A deep-learning time-series model integrated sound, images, and history to help creators navigate a large template marketplace.
+20% adoption+33% session duration
04Brand recommendation
BrandAlley
Fashion discovery through brand affinity.
Recommendation work that connects shopper preference with relevant brands across a changing fashion assortment.
05AI taxonomy
Digital Turbine
Custom categorization for app inventory.
High-volume content classification and ML systems designed to turn noisy signals into useful targeting and discovery structure.
Client results
Built with product teams. Measured in the product.
20%increase in template adoption
33%longer session duration
150%sales lift vs. previous model
$250kmonthly cost reduction
“Argmax took our search and recommendation capabilities to the next level. By integrating advanced matrix factorization and visually-driven reranking, they not only made our recommendation feeds more engaging but also significantly improved our play-to-download rate.”
Gili YogevAI & Data Science Team Lead, Artlist
“Partnering with Argmax transformed our approach to template recommendations at Lightricks. Their deep learning expertise enabled us to integrate sound, images, and historical data into a powerful solution, leading to a 20% increase in template adoption and 33% longer session durations.”
Shaked ZychlinskiHead of Recommendations, Data Science, Lightricks
“We hired Argmax to optimize our Real-Time Bidding offering, and their expertise made a significant impact. They helped us implement machine-learning-driven ad-user matching, allowing us to better measure and enhance profitability.”
Ofir PasternakFounder / R&D Director, Persona.ly
“Their AI-driven recommendations boosted sales by 150%, delivering targeted promotions that significantly enhanced engagement with small businesses. This solution provided impactful results and helped us refine our approach.”
Alex LevitSEE VMS Cluster Lead, Visa
Start with your inventory
Your marketplace is unique.
Your search stack should be, too.
Show us your listings, queries, and discovery bottlenecks. We’ll help you find the highest-impact place to start.
Talk to a marketplace search expert