Technical Program Manager Computer Vision, AI/ML & Mobile

Curious about the role?

Tiger Analytics is building Intelligent Product Engineering (iPE), the practice that owns end-to-end product engineering for our most complex client programs, anchored in an AI-native software development lifecycle. We are hiring a Technical Program Manager to run one of its most demanding programs.

Tiger Analytics is a global data and AI services firm working with large enterprises across retail, CPG, financial services, healthcare and technology. Intelligent Product Engineering is where that heritage meets full product and application delivery, with AI as the connective thread.

The program is a computer vision and machine learning product delivered into the hands of field users through a production mobile application. The application is built in React Native with a cross-platform bridge into native machine learning modules, and inference runs both on-device and server-side, each with its own constraints. The program spans model development and evaluation, that hybrid inference path, mobile and backend engineering, data pipelines, and rollout across markets where devices, connectivity, lighting and user behaviour all vary. The technology is genuinely new; the delivery problem is not, and it is unforgiving.

This is not a coordination role. You will own the plan, the requirements, the gates and the truth about status. The people you work with, architects, ML leads, mobile engineers, client product and technology stakeholders, are experienced and opinionated. The job is to hold the whole system in view, convert ambiguity into agreed and testable scope, and defend a plan with evidence, including when that means telling a client or a senior engineer that a date, a design or a team shape will not work.

This role sits on a program where the technology is genuinely novel and the stakes are real: a computer vision product that has to work in the field, on real devices, for users who will abandon it if it does not. The person who runs it well will have both a rare reference story and a direct hand in defining how iPE delivers AI-native products at scale.
Key Responsibilities:

Program Ownership

• Own the end-to-end plan: scope, sequencing, dependencies and critical path across ML, mobile, backend, data and QA workstreams that release on different clocks.
• Run the operating cadence, such as the planning, design and release gates, risk and issue management and maintain a single version of status that both Tiger and the client trust.
• Own release readiness. Define exit criteria in advance and hold them. Nothing ships because the date arrived.
• Forecast honestly: surface slippage the week it becomes visible, with options and a recommendation attached.

Requirements and Scope Control

• Convert ambiguous business intent into written requirements with testable acceptance criteria before build begins, working with client product owners and market stakeholders.
• Run change control that ties every scope addition to an explicit conversation about timeline, cost and team shape, every time, in writing.
• Maintain traceability from business outcome to requirement to test to release, so scope disputes are settled with a document rather than a recollection.
• Distinguish a genuine requirement gap from late discovery of something that was never specified, and be willing to say which it is.

Technical Judgement and Challenge

• Interrogate estimates, designs and team structures proposed by architects, ML leads and SMEs. Understand enough of the CV/ML lifecycle - data collection and annotation cost, training and evaluation, the gap between offline metrics and field accuracy, drift and retraining - to know when an estimate is thin.
• Push back on timelines, sequencing and staffing with evidence rather than opinion, and escalate cleanly when the answer is still no.
• Hold the hybrid inference path to account: parity and versioning between the on-device and server-side models, behaviour when the device is offline or the model is stale, and the latency, battery, binary-size and accuracy trade-offs that decide where a given inference should run.
• Treat engineering standards as delivery gates, not aspirations: branching and merge strategy across markets, CI/CD, automated test coverage on both the JavaScript and native layers, crash reporting, telemetry and observability, and validation of any configuration that can change behaviour in production.
• Drive diagnosis when the field reports th…