AI Engineer Agent & Runtime Engineering

AI Engineer — Agentic AI & Runtime Engineering

Company: SecNinjaz Technologies LLP

Experience: Around 2 years of relevant hands-on experience

Employment: Full-time, on-site

Locations: Netaji Subhash Place, New Delhi,

About the opportunity

At SecNinjaz, we are building an AI-powered Vulnerability Assessment and Penetration Testing (VAPT) product that helps organisations identify, investigate and validate security weaknesses in authorised environments.

We are looking for an AI Engineer who enjoys building agentic systems and turning working prototypes into reliable product capabilities. You will develop the agent workflows, runtime services, tool integrations and context systems that enable our product to perform complex cybersecurity tasks.

Strong AI engineering and software development skills are essential. Previous cybersecurity or VAPT experience is a plus. You will work alongside our cybersecurity specialists to understand the domain and translate their expertise into reusable AI capabilities.

What you will work on

• Agentic workflows: Build agents that plan tasks, select tools, interpret results and complete multi-step workflows with clear stopping conditions.
• Runtime engineering: Implement persistent state, checkpoints, retries, timeouts, cancellation and recovery so workflows remain dependable during failures and interruptions.
• Tool integration: Connect agents to security tools, APIs and internal services through structured interfaces, including Model Context Protocol (MCP) where appropriate.
• Reusable agent skills: Work with security researchers to convert tested procedures into versioned, documented skills that different agents and models can use.
• Context and retrieval: Build retrieval-augmented generation (RAG), context management and memory mechanisms that provide relevant knowledge and preserve the history of an investigation.
• Model integration: Integrate and compare LLMs through model adapters, with a focus on private and self-hosted inference.
• Execution controls: Implement scope restrictions, approval steps, permissions, resource limits and isolated tool execution in the application.
• Observability: Capture tool calls, evidence references, failures, latency and resource usage to make agent behaviour understandable and debuggable.
• Product improvement: Work with the data and evaluation engineer to diagnose failures, test changes and release improvements with regression checks and rollback support.

What we are looking for

• Strong Python programming skills and experience with APIs, backend services, Git and debugging.
• Hands-on experience building an LLM application or AI agent with tool calling, structured outputs and meaningful failure handling.
• Familiarity with an agent framework such as LangGraph, an agent SDK or an equivalent custom implementation.
• Understanding of RAG, embeddings, context windows and the limitations of LLM-generated responses.
• Working knowledge of asynchronous execution, databases and state management.
• Comfort working with Linux, Docker and application deployment.
• Ability to write maintainable code, test your work and explain your engineering decisions.
• Curiosity about cybersecurity and willingness to learn from domain specialists.

We value projects you can explain, modify and debug. Relevant professional work, research and substantial personal projects can demonstrate your capabilities.

Good to have

• Familiarity with VAPT, web/API security, authentication, authorisation or security testing tools.
• Experience with MCP, reusable agent skills or multi-agent coordination.
• Exposure to local model serving, vLLM, model gateways or GPU inference optimisation.
• Experience with PostgreSQL, vector databases, queues, tracing or CI/CD.
• Experience deploying AI applications in private or on-premises environments.

You do not need experience with every technology listed.

What you will gain

• Hands-on experience with SecNinjaz's on-premises NVIDIA H200 GPU infrastructure for product development, inference and experimentation.
• An opportunity to deepen your expertise in agentic AI, model serving and dependable AI systems.
• Direct collaboration with cybersecurity researchers and engineers on practical security problems.
• Ownership of product capabilities from implementation through evaluation and deployment.

Your initial contribution

Build an end-to-end agent workflow with useful execution traces, integrated tools and reliable recovery. Work with the security team to assess its results and with the evaluation engineer to measure improvements across releases.

How to apply

Apply through LinkedIn with your CV and one or two relevant project links, or a short technical write-up describing what you built and your contribution.

or Send email your CV to [Confidential Information] with details of your completed and ongoing AI/agentic AI projects, your specific contributions, current CTC, expected CTC, and notice period or earliest date you can join SecNinjaz.