"Technical Project Manager, AI Application Delivery
Duration: 12+ Months
Location: Remote
Work Authorization: H1B, EAD, USC
Position Summary
We are seeking a Technical Project Manager to lead delivery of AI-driven applications for internal business functions across [Quality / Manufacturing / Clinical / Regulatory / Commercial]. You will run an Agile team of engineers, data specialists, and business analysts building on Google Cloud, with Anthropic Claude and Google Gemini models served through Vertex AI.
This role is not administrative project tracking. You will make real tradeoff decisions about scope, model selection, cost, and validation sequencing, and you will be the person who reconciles an iterative delivery cadence with the change control and computer system validation obligations of a regulated pharmaceutical environment. That reconciliation is the hardest part of this job and the reason it exists.
You will partner closely with the Business Analyst, AI & Enterprise Data Products, who owns the requirement definition and model evaluation criteria.
You own delivery.
Key Responsibilities
Agile Delivery
Own end-to-end delivery for AI application initiatives: planning, execution, risk management, stakeholder communication, and release.
Facilitate Agile ceremonies — backlog refinement, sprint planning, daily standups, reviews, and retrospectives — and hold the team to working agreements.
Administer and continuously improve the team''s Jira configuration: project setup, workflows, issue types, custom fields, automation rules, and JQL-driven dashboards and reports.
Own the team''s Confluence space as the single source of truth: architecture decision records, meeting notes, runbooks, onboarding documentation, and links between requirements, tickets, and validation evidence.
Manage dependencies across platform, data engineering, security, infrastructure, and business teams; escalate and unblock rather than merely report.
Produce clear status communication calibrated to audience — an executive summary is not a sprint burndown.
AI-Specific Delivery Management
Plan and sequence work that is inherently uncertain. Structure model and data exploration as timeboxed spikes with explicit success and kill criteria, and resist committing model-quality outcomes to fixed sprint scope.
Manage model lifecycle as a first-class dependency: track Claude and Gemini model version releases and deprecation notices, plan migration windows, and ensure evaluation is re-run against any model change before it reaches production.
Own inference cost management. Forecast token consumption against projected adoption, monitor spend against budget, and drive architectural decisions — model routing, caching, prompt efficiency, context management — when unit economics require it.
Coordinate evaluation as a release gate. A build is not done because tests pass; it is done when it passes the agreed evaluation thresholds on the golden dataset.
Manage the operational realities of model-backed systems: rate limits, latency budgets, quota management, fallback behavior, and incident response when a model dependency degrades.
Facilitate model selection decisions between Claude and Gemini for specific workloads based on evaluated performance, cost, latency, and context requirements — and document the rationale for audit purposes.
AI-Assisted Software Development
Enable and govern the team''s use of AI-assisted development tooling (Claude Code, Gemini Code Assist, or equivalent) including productivity expectations, code review standards, and testing rigor when generation volume increases.
Partner with Security, Legal, and IP stakeholders to ensure AI-assisted development practices comply with company policy on source code handling, third-party data, and confidential information.
Track and report on where AI-assisted development is genuinely improving throughput versus where it is shifting bottlenecks to code review and quality assurance.
Google Cloud Platform Delivery
Coordinate delivery across Google Cloud Platform services relevant to the stack: Vertex AI, BigQuery, Cloud Run or GKE, Cloud Storage, Dataflow or Composer, and Document AI.
Partner with Cloud Engineering and Security on IAM, VPC Service Controls, CMEK, network architecture, and environment strategy (dev/test/prod).
Manage cloud consumption and cost forecasting alongside inference spend.
Coordinate supplier qualification and shared-responsibility documentation for cloud infrastructure supporting GxP-relevant systems.
Compliance & Validation Delivery
Plan and sequence validation activities so they enable rather than obstruct iterative delivery — establishing which changes require full revalidation, which are covered by an approved change plan, and which fall outside validated scope.
Work within the company''s CSV/CSA framework and GAMP 5 (Second Edition) principles, applying risk-based rigor proportional to patient safety, product quality, and data integrity impact.
Own the change control calendar and coordinate approvals with Quality Assurance ahead of release windows.
Ensure the audit trail from requirement through ticket, code, test evidence, and approval remains intact and inspection-ready.
Represent the delivery process to internal audit and health authority inspectors when required.
Team & Stakeholder Leadership
Build a working environment where engineers can raise risk early without penalty.
Manage vendor and system integrator relationships, including statements of work, deliverable acceptance, and performance.
Coach the team and business stakeholders on realistic AI capability, and manage expectations in both directions.
Contribute to resource planning, hiring, and onboarding for the team.
Required Qualifications
Bachelor’s degree in computer science, Information Systems, Engineering, or a related field, or equivalent practical experience.
5–8 years managing software delivery, including at least 3 years running Agile teams as Scrum Master, delivery lead, or technical project manager.
Demonstrated delivery of at least one production application involving machine learning, LLMs, or advanced analytics.
Advanced Jira and Confluence proficiency — able to configure workflows, build automation, write JQL, and structure documentation spaces without administrator support.
Working knowledge of a major cloud platform, with Google Cloud strongly preferred.
Sufficient technical fluency to challenge an engineering estimate, follow an architecture discussion, and understand the tradeoffs in a model selection decision.
Excellent written and verbal communication across engineering, business, and executive audiences.
Authorized to work in the United States without current or future sponsorship. (Adjust per company policy.)
Preferred Qualifications
Experience in pharmaceutical, biotech, medical device, or another FDA-regulated industry.
Direct experience delivering a GxP-relevant system, including change control and validation coordination with Quality Assurance.
Hands-on experience with LLM application delivery: prompt iteration, retrieval-augmented generation, evaluation frameworks, and observability tooling for model-backed systems.
Familiarity with Vertex AI Model Garden, including serving third-party models such as Claude alongside Gemini.
Experience managing cloud and inference cost at scale (FinOps practices).
Familiarity with the NIST AI Risk Management Framework and evolving FDA guidance on AI in regulatory decision-making.
Experience with Atlassian ecosystem extensions — Advanced Roadmaps, Jira Product Discovery, Compass, or Rovo.
Certification such as PMP, PMI-ACP, CSM/A-CSM, SAFe, or Google Cloud Digital Leader / Professional Cloud Architect.
Core Competencies
Decisiveness under uncertainty — makes and documents calls with incomplete information rather than waiting for clarity that will not arrive.
Technical credibility — earns engineers'' respect by understanding the work, not by tracking it.
Regulatory pragmatism — finds the path that satisfies Quality without stalling the team, and knows which battles are worth having.
Cost consciousness — treats inference and cloud spend as a design constraint, not an afterthought.
Candor — surfaces bad news early and accurately.
Ownership — follows issues to closure across organizational and vendor boundaries." This is JD share Portels short version
Never miss an opportunity! Create an alert based on the job you applied for.
