Staff Machine Learning Engineer, Multimodal Modeling

Flock is the leading safety technology platform, helping communities thrive by taking a proactive approach to crime prevention and security. As a Staff Machine Learning Engineer, you will lead the advancement of core embedding-based retrieval systems, focusing on fine-tuning and extending multimodal models to improve performance and ensure extensibility across evolving product use cases.

Responsibilities:

• Lead the advancement of core embedding-based retrieval systems with a primary focus on the scientific aspects of modeling
• Fine-tune and extend multimodal models (e.g., CLIP, SigLIP) to improve performance, generalization, and cross-modal alignment
• Work on unifying text and image representations, improving model performance, and ensuring extensibility across evolving product use cases
• Own long-term maintenance and become a leader for the team offering hands-on help
• Begin exploratory work

Requirements:

• 7+ years of industry experience in Machine Learning with a focus on representation learning, multimodal modeling, or embedding-based retrieval
• Deep domain knowledge in at least one area: computer vision, natural language processing, or recommendation systems
• Strong proficiency in PyTorch, with experience fine-tuning foundation models and adapting pretrained vision-language models to real-world tasks
• Demonstrated ability to customize and extend model architectures, training loops, loss functions, and data pipelines to deliver impact
• Experience with embedding-based retrieval, including contrastive learning, multimodal alignment, and designing evaluation methods for vector similarity search and embedding quality
• Solid engineering fundamentals in Python, with familiarity in Git, SQL, and Bash
• Comfortable working independently and navigating ambiguity, with a track record of solving open-ended modeling problems
• Ability to obtain and maintain Criminal Justice Information Services (CJIS) certification as a condition of employment
• Familiarity with model compression techniques, such as distillation, quantization, and architecture pruning, to improve inference efficiency and deployability
• Experience with vector search infrastructure, including provisioning, maintaining, and querying large-scale vector databases (e.g., FAISS, Weaviate, Pinecone)
• Proficient with multi-GPU and distributed training workflows, to scale training of large multimodal models efficiently