Technical Product Manager - Schedule Studio

Increasingly, the product itself is agents. Our direction is to move from tools that help people plan to the AI that does the planning - agents that read the scope, build the programme, spot what's wrong with it, and propose what to do, with the planner signing off rather than doing the legwork. You'll design these agents: what they should do, what they must never do, and where the planner stays in the loop when the consequences are measured in years and billions.

And because you can't ship an agent you can't measure, evals sit at the centre of how we work. For Schedule Studio that means deciding what a "good schedule" is in terms a machine can score - logic density, open ends, float distribution, milestone realism, resemblance to what experienced planners actually produce - and proving every change against it. Nothing counts as done until a structured evaluation shows it's actually better.

What you'll do

• Be the planner in the room: you are the standard for what Schedule Studio produces. Review generated programmes the way you'd review a contractor's submission, find what's wrong, and turn that into product and model improvements. Own the definition of schedule quality across the product.
• Own the squad's quarter: plan it, reconcile it against unplanned work and support demands, and deliver it. Make build/kill decisions on evidence, and know what value a piece of work will create - and how you'll measure it - before you build it.
• Build, not just brief: write clear specs, prompt agents effectively, and review and edit generated code and output rather than writing from scratch or handing over. Keep idea-to-test in days by choosing the cheapest path to a valid test.
• Design and ship planning agents: decide what an agent should do (build the programme, challenge the logic, propose a recovery sequence), what it must never do, and where the planner stays in the loop. Shape the context, tools, and guardrails it works with - including how it uses project data, templates, and the P6 export.
• Run evals - this is central to what we do: define what a good schedule looks like before you build, then prove it with structured evaluations - representative project sets, scoring rubrics a senior planner would agree with, and success metrics covering quality, cost, and reliability. A schedule that looks plausible but a planning manager would reject doesn't ship. Kill things fast when the evidence says so; we measure ourselves in outcomes shipped and experiments correctly killed.
• Get close to planners and the people they answer to: run user and buyer research with planning managers, project controls leads, and programme directors; sit in on client and sales meetings; and treat real feedback from working planners as your most valuable input. Our clients run multi-billion-dollar projects and carry real accountability for them; understanding their world is the job - and you already do.
• Lead contributors, not a process: lead your squad to decisions and delivery with no handovers, coordinate dependencies with the squads working on forecasting, portfolio, and our AI agent, and spread what your squad learns through our guilds (our cross-squad groups for frontend, backend, and AI practice).

Please mention the word 'float' in your application.

What we're looking for

• Five-plus years as a planner or scheduler on major construction or infrastructure projects - building, maintaining, and defending programmes that were contractually and commercially live, not just reporting on them.
• Deep fluency in Primavera P6 (Asta, MSP, or TILOS a bonus): logic, calendars, constraints, baselines, WBS and coding structures, and the many ways a schedule can look fine and be wrong. You can read a programme and tell within minutes whether you'd trust it.
• A clear point of view on schedule quality - you can articulate what makes a programme credible, know the common quality checks (and where they fall short), and have argued about float across the table and held your ground.
• You already build with AI, daily: you've used LLM tools to draft, analyse, or automate parts of your own work, and you want to go much further - to prototypes and working code you review and edit rather than write from scratch.
• Judgment over process: you can take an ambiguous problem, get to the heart of what planners and their clients actually need, and make the right call because you understand the trade-offs, not because you followed a framework.
• You're fluent with data: you use it to find the problem worth solving and to know whether what you shipped actually worked - and you're excited by the idea of defining schedule quality in a way that can be measured at scale.
• You communicate sharply: you can make a complex trade-off legible to an engineer, a planning manager, and a programme director in the same afternoon.
• You've changed your mind on evidence. You can point to a sequence, a plan, or an approach you believed in, what the data told you, and what you did about it.
• The problems are hard, the stakes are real, and the path isn't always clear - you're energised by that, not slowed down by it.

Nice to have

• Product experience of any kind - owning a feature area, a tool, or an internal system that other planners depended on - or experience building planning tooling, scripts, or automation for your team.
• Experience with data science, machine learning, or probabilistic risk analysis (QSRA/QCRA), and comfort talking about forecasting, uncertainty, and probabilities.
• Experience working across both client and contractor sides, or across multiple sectors (rail, energy, nuclear, water, buildings, oil and gas).
• Knowledge of contract programme requirements under NEC, JCT, FIDIC, or similar, and of what "acceptance" of a programme actually involves.
• Some formal technical background (engineering, development, or data).

**What working