Product Manager, Agent Harness & Modelling

About Cohere

Cohere is a leading security-focused enterprise AI company creating advanced foundational AI models and complete products crafted to address practical business challenges. We develop and deploy cutting-edge models to empower organizations building AI systems, aiming to drive widespread AI adoption globally. Our collaborative team of researchers, engineers, and designers passionately contributes to enhancing our models' capabilities and delivering value to customers. Headquartered in Toronto, our presence includes cities like London, New York City, San Francisco, Montreal, Paris, Berlin, and Seoul.

About the Role and North Platform

Cohere is transforming enterprise AI through North, our AI platform designed to securely integrate AI agents and automation within organizational infrastructures. North helps employees simplify workflows, automate routine tasks, and extract actionable insights while strictly protecting data privacy and compliance. It combines state-of-the-art generative and search models with flexible integrations to boost productivity and innovation at scale.

The Agent Harness Product Manager will oversee the execution framework that ensures North agents are dependable, functional, and ready for production use. This role intersects three key areas:

• Agent Loop and Execution: Direct ownership of the core agent runtime including tool orchestration, parallel task execution, sub-agent delegation, sandboxed code execution, and failure recovery. Defining how agents plan and operate over extensive multi-step processes, ensuring robustness for enterprise demands. Engagement at the implementation level alongside engineers is expected, not just requirement handoff.
• Context Engineering: Management of the context window as a controlled resource, including progressive tool/skill disclosure, context summarization and compaction, offloading large payloads to persistent storage, and maintaining agent orientation over extended tasks.
• Model-Scaffolding Integration: Serving as the liaison between the harness and Modeling teams to validate harness design choices before construction, using evaluations to bridge product and research areas, and ensuring the co-evolution of model and harness functionalities.

Key Responsibilities

• Lead the strategic roadmap development for North's agent harness comprising agent loop, context engineering, tool orchestration, sandboxing, and sub-agent delegation.
• Act as the primary liaison between North engineering and Cohere's Modeling team to validate harness features prior to implementation and prevent design dead ends.
• Manage North's agent evaluation framework, ensuring compatibility with both harness infrastructure and modeling training systems and facilitating trust between product and research teams.
• Collaborate with enterprise customers to identify agentic failures and translate real-world issues into actionable product and model requirements.
• Partner with the modeling team expert in agent systems to set impactful goals, prioritize model features, and develop a clear capability advancement plan with relevant evaluations.
• Keep abreast of developments in open-source and commercial agent ecosystems and guide adoption strategies that maintain alignment with evolving architectural standards.

Required Qualifications

• Over 5 years of product management experience focused on research PM roles, agentic AI systems, or applied machine learning products.
• Thorough expertise in agent harnesses, encompassing the latest commercial and open-source technologies.
• Comprehensive knowledge of modern large language model (LLM) agent architectures, including multi-agent frameworks, tool-enabled reasoning, memory and retrieval systems, programmatic orchestration, retrieval-augmented generation, and extended horizon execution.
• Proficient understanding of agent evaluation methods, including metrics for task completion, failure recovery, reliability over long durations, and isolating model versus scaffolding shortcomings.
• Technical aptitude to contribute to architectural decisions, review design documentation, and evaluate asynchronous execution, sandboxed environments, filesystem structures, and production platform trade-offs.
• Capable of fluidly transitioning between machine learning research discussions and engineering architecture conversations.
• Proven track record of delivering platform-layer products that significantly improve reliability, performance, or functionality.

Desirable Experience

• Experience in developing and deploying leading agent harness systems.
• Active engagement with agent frameworks, including coding agents and orchestration tools both professionally and personally, staying current with emerging harness technologies.
• Hands-on knowledge of enterprise agentic deployments, including multi-tenant orchestration, tool permission controls, audit logging, and compliance management.
• Understanding of infrastructure limitations relevant to enterprise contexts, such as on-premises deployment, scalability hurdles, and operational compromises in air-gapped or restricted settings.
• Experience balancing research insights with product delivery to translate preliminary model capabilities into launched features.
• Background in working closely with machine learning research or post-training teams.

Why Work at Cohere?

• Make an impact by influencing how key Canadian institutions adopt frontier AI technologies.
• Innovate alongside top researchers and engineers tackling complex machine learning problems.
• Benefit from competitive pay, equity participation, and professional growth opportunities.
• Enjoy flexible work arrangements with hybrid office options in multiple global cities including London.

Employee Benefits

• Weekly lunch allowance of $75 or £75 (or local equivalent).
• Comprehensive health and dental coverage with additional budget earmarked for mental health.
• Retirement plans including RRSP matching, 401K, or Pension schemes depending on location.
• Parental leave top-up at 100% for up to six months for either parent.
• Annual enrichment allowances for arts, culture, fitness, wellness, personal time, and workspace improvements.
• Dedicated education and learning stipends to support conferences, courses, and coaching.
• Generous paid vacation of six weeks (30 working days).
• Travel budget for visits to other company offices and annual company-wide retreats.

Work Environment and Culture

• Cohere supports remote work but also maintains offices in various global tech hubs including London.
• Office-based employees receive daily lunches, snacks, and participate in regular social and community events.
• Remote employees are supported with coworking benefits to facilitate local collaboration.
• All employees can claim a $500 stipend to optimize their home office setup.

Additional Information

Cohere encourages candidates with diverse backgrounds to apply even if not all listed qualifications are met, fostering an inclusive work environment. Accommodations during recruitment are available upon request. The recruitment process may include AI-assisted screening to identify qualified applicants but will not limit human review. Beware of fraudulent job offers; all legitimate communications come from official domains following Cohere's channels.

Compensation

The annual salary for this position ranges from $160,000 to $320,000.