Build Log

Infrastructure

• Provisioned high-performance Azure VM (Standard_D2ds_v4)
• Automated SSL provisioning via Let's Encrypt and Certbot
• Nginx reverse proxy for secure traffic routing

Deployment Pipeline

• Created sync-api-to-vm.sh for atomic code deployment
• Automated Docker Compose orchestration on remote server
• Standardized environment variable management

• Created public build log at /build-log
• Filter by status and category
• Shareable anchor links per entry
• SEO-optimized with meta tags

Capabilities

• Parses uploaded resumes via PDF extraction
• Generates one-paragraph candidate summary
• Identifies strengths and concerns
• Provides fit score from 0-100
• Makes recommendation: Strong fit, Potential, or Not suitable

Implementation

• Azure OpenAI integration with GPT-4o
• Async processing with background tasks
• Auto-screening on application submission
• Results cached for instant retrieval

How it works

• Each candidate gets a 1536-dimension embedding via text-embedding-3-small
• Queries converted to embeddings in real-time
• pgvector with HNSW index finds similar candidates instantly
• Results ranked by cosine similarity (0-100%)

Performance

• Sub-second response times
• Handles thousands of candidates
• No more missed matches due to keyword variations

Features

• SEO-optimized with proper meta tags
• Mobile-responsive design
• Custom questions per job
• Resume upload with drag-and-drop
• Progress autosave (never lose an application)
• GDPR-compliant consent checkbox

Planned

• Template library per workspace
• Variable substitution (candidate name, job title, etc.)
• Preview before sending
• Sent email history with status tracking

Capabilities

• Natural language questions about candidates
• Context-aware responses based on current view
• Conversation history
• Suggested follow-up questions

Planned metrics

• Time-to-hire by role
• Application source effectiveness
• Pipeline stage conversion rates
• AI screening accuracy over time