AI Prompt Engineer (Skills & Prompts, PhD Required)

Pixalate is seeking a talented AI Prompt Engineer to join our remote team based in Pakistan. In this role, you will be responsible for designing, writing, and maintaining the system prompts and agent skills that govern how our AI agents operate in real-world customer scenarios. These agents tackle complex challenges such as investigating invalid traffic (IVT) and providing direct data analytics through interactive AI. Your work will be critical in shaping agent behavior, ensuring they ask clarifying questions when needed, select appropriate tools, and deliver accurate results. This is a hands-on engineering position focused on building robust, production-ready AI systems rather than simple prompt experimentation. You will collaborate closely with teams developing key products like the IQ Agent, Analytics MCP, and client IVT investigator agents, and will be accountable for agent performance in live environments.

Key Responsibilities

- Build and maintain Claude skills and plugins that define agent capabilities and behaviors.
- Design agents to interactively gather requirements by asking clarifying questions instead of providing one-shot answers.
- Determine the appropriate distribution of logic between skills, system prompts, and tools.
- Continuously update and manage the skill library as products evolve, treating it as a living system.
- Own and version the system prompts for Pixalate’s client-facing agents, including IVT investigators and MCP products.
- Debug agent failures such as incorrect tool usage, hallucinated outputs, missed clarifications, and context limitations.
- Define success criteria for each agent and develop evaluation suites to measure performance.
- Create ground-truth datasets from real customer data to benchmark changes and prevent regressions.
- Engineer context strategies to optimize what agents see, in what sequence, and within resource budgets.
- Collaborate with engineering teams on MCP tool interfaces, ensuring accurate tool descriptions and selections.
- Optimize agent behavior for cost efficiency, latency, and accuracy.
- Safeguard against prompt injection, data leakage, and unsafe tool usage in production agents.
- Ensure agents represent Pixalate’s findings accurately without overstating confidence.

Required Qualifications

- PhD in Computer Science, AI, Machine Learning, or a related field, or an exceptional research track record.
- Proven experience building and shipping LLM agents to real users, beyond prototypes or demos.
- Hands-on expertise with agent skills, plugins, system prompts, or similar agent-behavior tooling.
- Strong evaluation discipline with the ability to design evaluation suites, build ground-truth datasets, and detect regressions.
- Excellent written English skills, as writing is a core part of the role.
- Proficiency in Python or TypeScript and ability to understand the underlying codebase.
- Solid understanding of tool-calling/function-calling mechanisms and agent tool selection.
- Awareness of prompt injection risks, personally identifiable information (PII) handling, and AI failure modes.
- Willingness to work some hours overlapping with U.S. time zones for team collaboration.
- Familiarity with ad-tech, digital advertising, or fraud detection concepts (programmatic, DSPs, SSPs, CTV, IVT) is strongly preferred. Candidates without this experience must be ready to learn the domain deeply within the first 90 days.

Preferred Qualifications and Benefits

- Experience with Model Context Protocol (MCP) servers or similar tool-integration standards.
- Knowledge of observability and tracing in LLM systems.
- Background in retrieval systems and understanding their failure modes.
- Data analysis skills, including SQL and working with large datasets.

At Pixalate, you will join a global team of lawyers, data scientists, engineers, economists, and researchers dedicated to protecting businesses, consumers, and children from deceptive and fraudulent digital environments. Our work has been recognized by UNICEF and featured in major media outlets for uncovering high-profile cases involving online privacy, ad fraud, and digital safety.

We foster a culture of small, high-output teams that value bold ideas, fearless attitudes, and data-driven collaboration. Communication is concise, and respect is earned through results rather than titles. Chess players and analytical thinkers thrive here.

What We Offer

- Competitive compensation package
- Monthly internet reimbursement
- Casual, fully remote work environment with flexible hours
- Opportunities for professional growth and advancement
- Engaging annual team events
- The chance to work alongside some of the brightest minds in AI and digital trust, impacting millions of users worldwide

If you are passionate about building cutting-edge AI systems that make a real difference and enjoy working in a fast-paced, collaborative environment, Pixalate is the place for you.