Technical Product Manager — Software and Agents (Core Platform)

Radical AI; New York (NY); On-site; USD 140,000 - 190,000; Full time

Job summary

Radical AI in New York seeks a Technical Product Manager to own the agentic and core software layers of our platform, turning agents, scientific models, data, and experimental workflows into a coherent product that drives customers from research goals to results.

You will define how users direct and inspect the system’s work, set priorities, and own the path from research prototype to production, defending the product position when needed.

Qualifications

• 5–7 years in product management, incl. at least 1 year on agents or ML.
• Technical credibility: reads code, prototypes, and reasons about architectures.
• Ability to balance demos with dependable agents and human-in-the-loop.
• Capable of tying offline metrics to real lab/customer outcomes.
• Turns agents, models, UX, and platform foundations into a cohesive product.
• Influences across teams without formal authority.
• First principles thinking to structure undefined problems.

Responsibilities

• Own the roadmap for agentic and core software layers of the platform.
• Define what capabilities are useful and reliable for production deployments.
• Collaborate with software engineers, ML researchers, and scientists.
• Direct and inspect user goals, agent plans, and outcomes.
• Balance research prototypes with production-ready features.

Skills

Product management

Technical credibility

Agent/ML experience

Measurement & evaluation

Product synthesis

Influence without authority

First principles thinking

Tools

APIs

DevOps

MLOps

Job description

We are looking for a Technical Product Manager to own the agentic and core software and infrastructure layers of our core platform. Your job is to turn agents, scientific models, data, and experimental workflows into a coherent product that takes customers from research goals to results. That means defining how users direct and inspect the system's work, how agents plan and execute research, and how the platform learns from each experiment to deliver better decisions and better outcomes.

There is no established PM playbook for this role, and we are not looking for someone to follow one. You will report to our Co-founder and President and work directly with software engineers, ML researchers, and materials scientists. Those conversations are substantive and sometimes contentious. This is a team with strong opinions and the technical depth to back them up.

This is not a coordination role. You will own the roadmap of our agentic and core software layers: what agents should be able to do, how users direct and inspect their work, and how the platform's data, models, and workflows come together as a coherent product. You will set priorities, define what makes a capability useful and reliable enough to deploy, and own the path from research prototype to production. That means defending a product position when engineers and researchers push back, knowing when to stand firm and when to update your view, and earning credibility through judgment rather than authority. If your instinct is to manage a backlog rather than own an outcome, or to defer rather than decide, this is not the right fit.

What You'll Own

The Core Platform:

• The core platform roadmap: the shared software, data, and infrastructure that every agent, model, and Workcell depends on. You decide what belongs in the core, what stays at the edges, and what we build versus adopt.
• Core services and APIs: workflow orchestration, experiment and sample records, data access, and the contracts that let agents, models, and the lab work together. You define them so teams build on the core instead of around it.
• The data foundation: how lab, simulation, and literature data are captured, structured, versioned, and made usable by agents, models, and scientists.
• Reliability and scale: performance, observability, security, and the bar a capability must clear before it becomes part of the core.

Agents on the Core:

• What agents can do on the platform: the tools and services they can call, how they plan and execute research, and where a human approves or steps in.
• How users direct and inspect the system's work: setting research goals, reviewing agent plans and decisions, and tracing any result back to its data, model, and experiment.
• Evaluation: the metrics, benchmarks, and acceptance criteria that show whether an agent or model improves real lab outcomes, not just demo or offline performance.

Models and the Learning Loop:

• How the platform serves, versions, and monitors the predictive and generative models that drive experiment selection, including when to build, fine-tune, or adopt.
• The learning loop: how model outputs feed experiment design and how results flow back into the core to improve the next decision. Where the loop is slow, noisy, or broken, you find it and fix it.

Customers and Governance:

• The interfaces through which our scientists and customers use the platform and consume results, such as APIs, data exports, and reports.
• Data governance: customer data separation, data rights, and handling of export-controlled information.
• Customer commitments: what we promise, how we measure it, and what we do not claim.

Cross-Functional:

• Work closely with the software engineering, ML research, and science teams to keep the roadmap grounded in both technical reality and customer need. The Technical PM (Robotics and Autonomy) owns the physical lab. You own the core platform that directs it and learns from it. Together you own the interface between the two.
• Collaborate with PMs at partner and customer organizations to align on data formats, requirements, and priorities.
• Communicate product direction clearly, both to the engineers and researchers building it and to leadership deciding where to invest.

What We're Looking For

Required:

• Experience. 5-7 years in product management, including at least 1 year on products built on agents or ML, ideally at a technical or deep-tech company.
• Technical credibility. You read code, build lightweight prototypes, and reason about agent architectures, APIs, data flows, and integrations. You work through feasibility and tradeoffs directly with engineers and researchers, without a translator and without becoming the implementation owner.
• Agent judgment. You know the gap between an impressive demo and a dependable agent: unreliable tool use, compounding errors, weak grounding, and misleading evaluations. You decide where autonomy adds value, where simpler software is better, and where a human belongs in the loop.
• Measurement. You tie offline model and agent metrics to real lab and customer outcomes, and you can tell when a benchmark is misleading.
• Product synthesis. You turn agents, models, user experience, and platform foundations into one coherent product rather than separate backlogs.
• Influence. You set priorities and earn alignment across ML research, science, design, and engineering, without authority and when those teams disagree.
• First principles. You reason from first principles and build structure where none exists.

Nice to Have:

• Experience with scientific software, such as materials, chemistry, or physics-based simulation, or surrogate models for expensive computation.
• Experience with Bayesian optimization, active learning, or closed-loop experimentation.
• Experience with DevOps and MLOps or data platform products.
• Exposure to enterprise customers.

Compensation and Benefits

Compensation varies based on experience and scope. We offer 100% employer-covered medical, dental, and vision insurance, unlimited PTO, 14+ company holidays, 2 weeks of company shutdown at EOY, 401(k), and mental health support.

Radical AI is an equal opportunity employer. We do not discriminate on the basis of race, color, ancestry, national origin, religion, sex, age, sexual orientation, gender identity and expression, marital status, disability, or veteran status.