AI Engineer – SaaS Platform

• Maintain and optimize LLM- and VLM-powered services for content generation, compliance scoring, and campaign testing
• Manage and scale Flask/FastAPI microservices, ensuring high uptime and low latency
• Maintain Dramatiq queues for asynchronous AI workflows, campaign generation, and pipeline orchestration
• Deploy, monitor, and debug Uvicorn/Gunicorn-based hosting in production environments
• Integrate with OpenRouter and equivalent LLM routing tools to balance cost, latency, and quality
• Design and refine prompt engineering strategies for reliability, context-awareness, and compliance
• Build and maintain feedback pipelines for AI model evaluation, including human-in-the-loop scoring, automated quality checks, and reinforcement
• Expose and maintain secure, versioned REST APIs for AI services
• Collaborate with backend and frontend teams to keep the microservice architecture aligned and maintainable
• Track token consumption, latency, and error rates to ensure production-grade performance

• Strong Python programming experience in production-grade codebases
• Experience with Flask; FastAPI optional
• Experience with Uvicorn/Gunicorn for async hosting
• Experience with Dramatiq, Celery, or RQ for background jobs
• Hands-on experience with LLMs and VLMs, including prompt engineering, fine-tuning, and evaluation
• Familiarity with OpenRouter or equivalent LLM/VLM routing and fallback tools
• Experience designing and maintaining microservice architectures
• Strong REST API design experience, including authentication, rate limiting, and documentation
• Experience with Dockerized deployments, CI/CD pipelines, logging/monitoring, and error handling
• Experience building structured evaluation and feedback systems for AI model performance
• AWS/GCP experience preferred
• 3–5 years as an AI Engineer or Python Backend Engineer working with production systems
• Prior work with SaaS platforms, LLM/VLM integrations, or AI-first products highly valued
• Demonstrated ability to maintain AI pipelines in production, not just prototypes