Lead AI Systems Architect, Data-Centric AI, Multimodal

This is a fully remote position, open to applicants in Sweden.

📋 Description

• Create and develop multimodal AI pipelines that integrate computer vision, speech-to-text (Whisper), and LLMs into seamless automated ingestion workflows.

• Implement frameworks for programmatic extraction (Instructor, Pydantic, DSPy, Outlines) to ensure strict and validated JSON outputs from generative models.

• Shift workloads from commercial APIs to optimized open-source models (Llama 3/4, Qwen, DeepSeek) operating on local inference engines (vLLM, TGI, Groq).

• Deploy data-focused AI frameworks (Cleanlab, Snorkel, Active Learning) to automate the labeling of millions of entries and to refine noisy datasets.

• Mentor junior developers, establish engineering best practices, and create architectural roadmaps for data enhancement.

• Transform raw, unstructured social media content (short-form video, audio, captions, and visual metadata) into precise, structured data products.

⛳️ Requirements

• Over 5 years of experience in Applied AI, Natural Language Processing, or Computer Vision.

• Demonstrated expertise in building multimodal pipelines that integrate vision, speech, and text models.

• Proficiency in Python, PyTorch/TensorFlow, and production-level LLM orchestration tools (DSPy, Instructor, Pydantic, LangChain/LlamaIndex).

• Practical experience in fine-tuning and deploying open-source models using vLLM, TGI, or other similar high-throughput inference engines.

• Familiarity with vector databases (Qdrant, pgvector, Pinecone) and RAG architectures at scale.

• Preferred: Previous experience with social media data pipelines (TikTok, Instagram Reels, YouTube Shorts).

• Preferred: Background in data-centric AI, active learning, or synthetic data generation.

🏝️ Benefits

• Flexible remote work arrangement.