ML Engineer, Agentic Video & Multimodal AI

Energy Jobline ZR; New York (NY); On-site; USD 175,000 - 300,000; Full time

Benefits offered by this job

Health plans

401K match

Commuter benefits

Catered lunch

Dinner stipend

Grubhub access

Wellness perks

Team offsites

Generous PTO

Job summary

Mirage is an AI- video platform seeking an ML Engineer to push the boundaries of large models for multimodal creative tasks. You'll develop new approaches for building and extending agentic systems that understand and operate over complex, real-world data, particularly video.

This role focuses on advancing agent capabilities, improving reasoning and control, and enabling new forms of interaction between models and time-based media.

Qualifications

• BS/MS/PhD in CS, ML or related field.
• Proven experience building production ML systems.
• Strong knowledge of transformers and LLMs.
• Experience with fine-tuning and alignment.
• Experience deploying agent systems.

Responsibilities

• Design and build end-to-end agentic systems for creative tasks.
• Develop novel training/adaptation approaches for large models powering agents.
• Design objectives, datasets, and fine-tuning strategies to improve agent behavior and reliability.
• Explore multimodal reasoning and structured for creative control.
• Run systematic experiments to evaluate and improve agent performance.
• Design evaluation frameworks for agentic workflows in video analysis and editing.
• Analyze failure modes across the full agent loop and iterate on improvements.

Skills

CS/ML degree

Production ML systems

Transformer models

Fine-tuning methods

Agentic pipelines

Full-stack ownership

Education

CS/ML degree

Tools

PyTorch

LLM tooling

Job description

Mirage is an AI- video platform seeking an ML Engineer to push the boundaries of large models for multimodal creative tasks. You'll develop new approaches for building and extending agentic systems that understand and operate over complex, real-world data, particularly video.

This role focuses on advancing agent capabilities, improving reasoning and control, and enabling new forms of interaction between models and time-based media.