Agentic Systems Engineer

Role Purpose
The Pakistan Digital Authority (PDA) is seeking an Agentic Systems Engineer to design, develop and test AI agents for government use cases. The role combines software engineering, retrieval and tool integration to deliver dependable applications that complete user tasks within defined permission and oversight controls.

Key Responsibilities

• Translate user needs into agent workflows, selecting single-agent or multi-agent designs according to task complexity and reliability needs.
• Build agent applications using frameworks such as LangGraph or model-provider SDKs.
• Integrate agents with authorised tools, APIs and databases, using Model Context Protocol (MCP) where appropriate.
• Develop retrieval-augmented generation (RAG) pipelines using vector and keyword search, with access-aware retrieval and source references.
• Write, version and test prompts, tool descriptions and structured outputs.
• Implement least-privilege tool access, input and output checks, and human approval for high-risk actions.
• Manage agent memory, context length and sensitive information, including retention and isolation between users.
• Create task evaluation sets and measure task success, tool-call accuracy, execution steps, latency and cost.
• Identify and resolve incorrect tool calls, loops and unreliable outputs; apply execution limits, timeouts and recovery controls.
• Work with the MLOps engineer to trace agent steps, investigate failures and support controlled production releases.

Required Qualifications & Experience

• Bachelor’s degree in Computer Science, Software Engineering, Artificial Intelligence, Data Science, or a related discipline.
• At least 4 years of relevant experience in software engineering or AI application development, including at least 1 year building LLM-based applications within that period.
• Demonstrated hands-on experience delivering applications that use tool calling, RAG or agent workflows.
• Proficiency in Python or TypeScript, API integration, databases, Git and software testing.
• Practical understanding of prompt design, retrieval, agent state and memory, structured outputs and application evaluation.
• Ability to implement permission controls and human approval workflows, with strong debugging, documentation and collaboration skills.

Preferred Experience

• Experience with LangGraph or comparable frameworks, MCP integrations and hybrid search systems.
• Experience addressing prompt injection, unsafe tool execution and unauthorised data access in AI applications.
• Exposure to government workflows, multilingual applications or enterprise systems requiring audit trails and human oversight.