AI Integration Engineer — Multimodal AI Platform - Freelance

**Location:** Fully remote (Canada) **Duration:** Milestone-based contract **Overview** A large-scale organization is building a next-generation multimodal AI platform to process vast volumes of imagery, video, and audio. The platform will perform first-pass analysis using computer vision, speech-to-text, and language models, then route findings to human experts for review and approval. This is a software architecture, integration, and data-engineering effort that assembles best-in-class existing AI services into one robust production platform — not a model-building project. The role requires a senior, hands-on engineer who has shipped similar systems and can make AI models swappable, observable, and production-grade. **Key Responsibilities** - Build integrations for vision, speech-to-text, and LLM services behind clean, standardized, swappable interfaces - Design and implement orchestration for sequential and parallel multimodal workflows, including running multiple providers for the same task and hot-swapping models with configuration changes - Implement Model Context Protocol (MCP) support so external agents and tools can plug in - Wire AI output (metadata, transcripts, summaries, confidence scores) into the human review flow - Capture workflow status, metrics, and evaluation signals - Design vendor-agnostic abstraction layers so underlying models can be replaced without touching the application **Required Qualifications** - 6+ years in software engineering with a track record integrating AI/ML services via APIs into production - Hands-on LLM integration experience (RAG, function/tool calling, orchestration frameworks) - Computer-vision or speech-to-text integration experience, ideally both in one system - Experience designing vendor-agnostic abstraction layers for model replacement - Strong Python skills and proficiency with relevant AI SDKs - Fluency across cloud AI providers - Portfolio or references demonstrating prior work that closely mirrors multimodal-integration scope **Preferred Qualifications** - Experience with Model Context Protocol (MCP) - Experience with evaluation/MLOps tooling and automated model-quality checks - Experience with cloud AI services (AWS Rekognition/Transcribe/Bedrock, Azure AI, GCP Vertex/Vision/Speech) - Experience building retrieval pipelines over large document or media stores - Experience delivering complex software systems for government or regulated industries **Work Location & Requirements** - Fully remote position - Must be based in Canada - Must be able to contract as an independent professional in Canada - Freelance/independent contractor engagement (not employment) - Selection process includes review of background and past work, followed by a short technical assessment/coding exercise