Stylitics is an AI-powered styling platform for top global retailers.

Services we provided:

Back-end development

Technologies used:

Clojure

About Stylistics

Stylitics is an AI-powered outfitting and styling technology partner for leading global retailers. Their platform transforms product catalogs into personalized, stylized outfit recommendations—enhancing customer engagement, boosting sales, and replacing manual styling processes with scalable automation.

> Go to website

Project Goals

Stylitics needed to automate their manual item categorization workflow, which was previously handled by a team in India. Every night, new retail items (clothing, accessories) were imported and required analysis for:

Type (e.g., dress, shoe)
Color
Style attributes (e.g., "bohemian," "formal")
Compatibility with other items for outfit generation

Our challenge

Our challenge was to replace human-driven classification with ML pipelines while ensuring:

Accuracy matching or exceeding manual efforts
Scalability for high-volume item processing
Seamless integration with Stylitics’ existing systems

Application Architecture

We designed a hybrid Clojure/Python stack to balance performance, maintainability, and ML capabilities:

Key Components

Data Ingestion Layer

Processed nightly batches of new retail items via event-driven workflows.
Integrated with Stylitics’ product databases using REST APIs.

Machine Learning Pipelines (Python)

Computer Vision Models: Analyzed item images for color, pattern, and style.
Text Processing: Extracted metadata from product descriptions.
Classification Engine: Tagged items with categories and outfit-compatibility scores.

Backend Services (Clojure)

Orchestrated ML jobs with fault-tolerant queues.
Stored processed data in PostgreSQL for real-time retrieval.
Exposed results via internal APIs to Stylitics’ frontend and analytics tools.

Legacy System Modernization

Migrated a critical sub-module from Clojure to Svelte for UI consistency and performance gains.
Added TypeScript, Cypress E2E tests, and Vitest to stabilize feature development.

Environments & Deployment

Local/CI Pipelines: Dockerized Clojure/Python services for parity across dev/staging/prod.
ML Model Training: Ran on GPU-optimized cloud instances (AWS SageMaker).
Hybrid Cloud: Deployed backend services to Kubernetes for scalability, with Python ML workers on autoscaling clusters.

Technologies

Backend & Data

Clojure (primary):
Duct (application framework)
HugSQL (PostgreSQL interface)
kafka-clj (event streaming)

Python:

PyTorch (image classification)
spaCy (NLP for product descriptions)
Celery (async task queues)

Frontend & QA

Svelte (migrated UI components)
TypeScript + ESLint (code quality)
Cypress (end-to-end testing)
Robot Framework (legacy test automation)

DevOps

Kubernetes (orchestration)
Terraform (infrastructure-as-code)
Prometheus/Grafana (monitoring)

Results

90% reduction in manual categorization work.
2x faster item processing vs. the human team.
Zero downtime during legacy-to-Svelte migration.
100% test coverage for critical paths post-refactor.
Stylitics now delivers real-time, AI-driven styling recommendations at scale—proving that Clojure/Python hybrids can power cutting-edge retail tech.