Inspiration

Skincare is deeply personal, yet most people receive generic advice. We wanted to build an AI advisor that analyzes your actual skin condition and delivers personalized recommendations — powered by cutting-edge AI infrastructure.

What We Built

Perfect Skin Adviser is an AI-powered skincare analysis and recommendation web app:

  1. Skin Analysis — Upload a photo, Perfect Corp's YouCam AI analyzes pore size, oiliness, age spots, and eye bags
  2. AI Recommendations — Nvidia Nemotron-3-Super-120B generates personalized skincare routines based on your results
  3. Resilient Infrastructure — TrueFoundry AI Gateway routes inference with automatic failover across providers

How We Built It

  • Frontend: Streamlit web app
  • Skin Analysis: Perfect Corp YouCam API
  • Agent Framework: Hermes Agent running on VPS, powered by Nvidia Nemotron on Crusoe Cloud Managed Inference
  • LLM Routing: TrueFoundry AI Gateway (Priority 0: Crusoe Nemotron → Priority 1: OpenAI fallback)
  • Resilience: Real-time provider status monitoring with automatic failover

Challenges

  • Integrating Perfect Corp's multi-step async API with real-time UI updates
  • Configuring TrueFoundry Virtual Models with priority-based routing to Crusoe Cloud
  • Making Hermes agent work with custom Crusoe inference endpoint

What We Learned

  • How enterprise AI gateways handle LLM failover in production
  • Nvidia Nemotron's capabilities for domain-specific recommendations
  • Building resilient multi-provider AI systems

Built With

  • crusoe-cloud
  • hermes-agent
  • nvidia-nemotron
  • perfect-corp-youcam-api
  • python
  • streamlit
  • truefoundry-ai-gateway
Share this project:

Updates