Applied AI Engineer - Manager

What matters most

• Grow engineers through meaningful ownership

  The role calls engineer growth its primary output. It specifies 6–12-month development plans, substantial early pairing and customer-meeting shadowing, timely feedback, and bi-weekly one-on-ones. For preparation, select a concrete example of how your management changed what an engineer could own.

• Own delivery while remaining hands-on

  The manager is accountable for technical outcomes across delegated workstreams and is expected to write code, review designs, pair on difficult work, and step in when a launch is at risk. Prepare to explain how you stayed close enough to detect delivery risks without taking ownership away from engineers.

• Make defensible technical decisions

  The role makes consequential architecture calls with incomplete information and sets standards for testing, review, code quality, and documentation. It also requires fluency in agents, tools, context, and evaluations. Prepare a production design example that makes assumptions, trade-offs, and likely failure points explicit.

• Earn customer trust through usable production systems

  Embedded engineers work with customer operators and executives; the manager owns the technical relationship. The posting emphasizes consistent delivery, clear communication when something breaks, and integration that does not disrupt users’ workflows. Prepare an example that connects an engineering decision to adoption or an operational outcome.

• Turn deployment lessons into reusable improvements

  The manager is expected to surface patterns from customer work and push reusable solutions into Mosaic, Percepta’s agentic-workflow toolkit. Prepare to distinguish a customer-specific fix from an improvement that would help a later deployment.

Likely challenges to think through

• Balancing coding and coaching

  Likely challenge, not a confirmed problem: rollout demands may pull the manager toward the code just when engineers need guidance and ownership. The posting says that balance shifts week to week. Prepare an example of deciding when to pair, delegate, or intervene directly, and how you protected development time.

• Deciding quickly in unfamiliar enterprise systems

  Likely challenge, not a confirmed problem: incomplete knowledge and launch pressure can make a fast architecture choice expensive to reverse. Prepare to walk through how you investigated an unfamiliar integration, recorded assumptions, chose what to test first, and communicated the remaining uncertainty.

• Reconciling local delivery with reusable tooling

  Likely challenge, not a confirmed problem: a tailored solution may serve one customer sooner, while generalizing it for Mosaic takes additional effort. Prepare an example of how you would judge whether a field pattern is reusable without delaying the immediate customer outcome or disrupting their workflow.

Topics to prepare

• Engineer growth and player-coach leadership

  The role centers on improving what engineers can own over 6–12 months while remaining active in the codebase. Review how to choose meaningful stretch assignments, pair during onboarding, and give timely feedback. Exercise: outline a development plan for a hypothetical engineer, including an assignment, early pairing, bi-weekly 1:1s, and observable signs of progress.

• Architecture decisions under ambiguity

  Percepta expects the manager to make expensive-to-reverse technical calls, explain trade-offs, and document assumptions. Exercise: compare two designs for a production AI system in an unfamiliar enterprise environment; identify what would strain first, what you would verify, and what decision you would make with incomplete information.

• LLM engineering for production workflows

  The posting explicitly calls for fluency with agents, tools, context, evaluations, and AI development tooling. Exercise: sketch an agentic workflow, then explain its tool boundaries, context choices, likely failure cases, and how evaluations would detect them. This is a preparation exercise, not a stated design assignment.

• Technical ownership across workstreams

  The manager remains accountable for delegated work, reviews enough code and designs to know what is shipping, and steps in when launches are at risk. Exercise: walk through a hypothetical at-risk rollout and decide what to inspect, what to delegate, when to intervene, and how to communicate the decision.

• Customer workflows and reusable deployment lessons

  Applied AI Engineers embed with customers; the manager owns the technical relationship and feeds reusable field insights into Mosaic. Exercise: map how operators would adopt a hypothetical AI workflow, identify disruption and integration risks, and distinguish a customer-specific fix from a reusable toolkit improvement. Enterprise integration experience is listed as preferred, not required.

Skills to revise

• Write and explain Python and TypeScript

  Both languages are explicitly named as Percepta’s stack and part of its interview loop. Exercise: implement a small, testable workflow component in each language and explain error handling, interfaces, and trade-offs; no particular framework is specified in the posting.

• Review code and define quality standards

  The manager sets standards for testing, review rigor, code quality, and documentation. Exercise: review a sample change, write specific review comments, identify missing tests, and draft a short definition of what must be documented before it ships.

• Construct an AI evaluation plan

  Evaluations are an explicit part of the required LLM fluency. Exercise: for a proposed agent-and-tool workflow, create representative success and failure cases, state what each checks, and describe how a failing result would change the implementation.

• Document an architecture call

  Percepta asks the manager to record the why behind consequential decisions, including assumptions made with incomplete information. Exercise: write a one-page decision record covering alternatives, trade-offs, assumptions, decision owner, and conditions that would prompt reconsideration.

• Deliver specific, constructive feedback

  The posting emphasizes immediate, respectful feedback and asks for a concrete example of an engineer whose capabilities improved through management. Exercise: rehearse feedback on a weak design, then prepare a factual account of an engineer’s starting point, the support provided, and what changed; do not claim an outcome that did not occur.

• Communicate a customer-facing issue

  The manager owns the technical customer relationship and must communicate clearly with nontechnical stakeholders, leading with the answer and then caveats. Exercise: give a brief status update for a hypothetical broken deployment: state the impact and next action first, then explain uncertainty and the next update.

• Map evidence to the role

  Prepare a one-page inventory of production software work, engineer leadership, and recent Python and TypeScript coding. The posting asks for 7+ years building production software and at least 2 years leading or managing engineers; technical leads with equivalent experience are welcome.

• Build a production-system walkthrough

  If you have a relevant example, outline the problem, your hands-on contribution, team ownership, customer workflow, launch risks, and outcome. An enterprise deployment is preferred experience, not a requirement.

• Practice coding in both languages

  Work through a small Python exercise and a small TypeScript exercise, then explain your tests and review choices. The posting says both languages are used in Percepta’s stack and interview loop.

• Draft an architecture decision

  As a preparation exercise, sketch an agentic workflow for an unfamiliar enterprise system. Record assumptions, tool and context choices, evaluation approach, integration boundaries, failure modes, and trade-offs you would revisit.

• Document an engineer’s growth

  Prepare a specific example, if available, of an engineer whose capabilities improved through your leadership. Show what changed and how assignments, pairing, onboarding, timely feedback, and ownership contributed.

• Rehearse a customer-facing incident explanation

  Prepare a brief answer-first account of a delivery problem: impact on operators, immediate action, uncertainty, and next update. This practices the direct technical and nontechnical communication the role calls for.

• Identify a reusable deployment lesson

  From an example you can discuss, distinguish a customer-specific fix from a pattern worth sharing across deployments. Explain how you would validate it before proposing it for Mosaic, Percepta’s agentic-workflow toolkit.

• Write questions for the employer

  Ask how the manager balances coding with team development during rollouts, how the Engagement Lead and Growth Leads divide decisions, and how customer findings become Mosaic improvements. Clarify customer-site expectations if relevant.