SecNinjaz Technologies LLP
Website:
secninjaz.com
Job details:
AI/ML Engineer — Agentic AI & Cybersecurity
Company: SecNinjaz Technologies LLP
Experience: 2+ years of relevant hands-on AI/ML engineering experience
Employment: Full-time, on-site
Locations: Netaji Subhash Place, New Delhi,
About the role
At SecNinjaz, we are building an AI-powered cybersecurity product that helps organisations identify, investigate and validate vulnerabilities in authorised environments.
We are looking for an AI/ML Engineer who enjoys developing intelligent agents, working with data and improving model performance through evaluation and fine-tuning. You will help build multi-agent workflows that use tools, interpret results and assist cybersecurity specialists with complex security tasks.
You will work alongside AI engineers, backend developers and security researchers to take capabilities from experimentation into the product. We welcome candidates with particular strength in agent development or model improvement and practical experience across both.
Strong AI/ML engineering skills are essential. Cybersecurity or VAPT experience is a plus, not a mandatory requirement.
What you will work on
- Agentic AI and multi-agent systems: Build workflows in which specialised agents coordinate tasks, call tools, exchange structured information and operate with clear objectives and stopping conditions.
- Agent runtime and integrations: Implement context management, persistent state, checkpoints, retries, timeouts and recovery. Integrate APIs, security tools and internal services through structured interfaces, including MCP where appropriate.
- Reusable agent skills: Work with security specialists to turn reviewed procedures into documented, versioned skills that agents and engineers can reuse.
- Data pipelines and quality: Convert approved agent traces, tool outputs, expert feedback and verified outcomes into structured datasets. Implement cleaning, deduplication, annotation, provenance tracking and sensitive-data handling.
- Model and agent evaluation: Build repeatable evaluations for task completion, tool-use accuracy, evidence quality, false positives, latency and compute cost. Maintain separate training, validation and held-out test sets.
- Fine-tuning and model improvement: Run reproducible experiments using supervised fine-tuning, LoRA, QLoRA or other suitable approaches. Compare candidate models against a baseline on unseen cases.
- RAG and retrieval: Improve how agents retrieve and use knowledge through evaluation of embeddings, rerankers, retrieval pipelines and context selection.
- Failure analysis: Investigate whether failures originate from the model, data, prompts, retrieval, tools or workflow design, then implement and measure the appropriate improvement.
- Product integration: Collaborate with the engineering team to integrate accepted changes with compatibility tests, regression checks, version tracking and rollback support.
- Execution controls: Help implement authorised scope, permissions, human approval steps and resource limits within the application.
What we are looking for
- Strong Python programming skills, including APIs, data processing, debugging and automated testing.
- Hands-on experience building an AI agent or multi-agent workflow with tool calling, structured outputs and meaningful failure handling.
- Familiarity with LangGraph, agent SDKs or equivalent custom implementations.
- Practical experience with PyTorch, the Hugging Face ecosystem or comparable AI/ML tools.
- At least one implemented LLM fine-tuning or post-training project covering data preparation, a baseline, held-out evaluation and failure analysis.
- Understanding of training versus inference, overfitting, generalisation, loss functions and evaluation metrics.
- Familiarity with dataset annotation, deduplication and prevention of leakage between training and evaluation data.
- Understanding of RAG, embeddings, context management and LLM limitations.
- Comfort working with Linux, Git, Docker and GPU environments.
- Ability to document dataset versions, configurations, checkpoints and results so experiments can be reproduced.
- Curiosity, clear communication and the ability to explain your own contribution and engineering decisions.
Relevant professional work, research, substantial personal projects and well-executed coursework can demonstrate these skills. Be prepared to walk through and debug something you have personally built.
Good to have
- Familiarity with cybersecurity, VAPT, web/API security, vulnerability reports or security testing tools.
- Experience evaluating agent trajectories, tool selection and multi-step task completion.
- Experience creating reusable agent skills or MCP integrations.
- Exposure to preference optimisation, DPO, reinforcement learning or distillation.
- Experience with model quantisation, GPU memory optimisation or distributed training.
- Familiarity with private model serving, vLLM, experiment tracking and on-premises deployment.
Cybersecurity certifications and previous H200 experience are not mandatory. You do not need experience with every technology listed.
What you will gain
- Hands-on opportunities to work with SecNinjaz’s on-premises NVIDIA H200 GPU infrastructure for planned training, fine-tuning, evaluation and inference workloads.
- Collaboration with cybersecurity specialists on practical security problems.
- Opportunities to deepen your expertise in multi-agent systems, data engineering and model improvement.
- Ownership of defined product features, supported by the wider engineering team.
- Experience taking AI capabilities from a working experiment to a tested product release.
Your initial contribution
Working with the team, deliver a bounded agent workflow with useful execution traces, establish a reviewed dataset and evaluation baseline, and complete a measurable improvement experiment. Explain the results on unseen cases, any regressions and the remaining limitations.
How to apply
LinkedIn Easy apply
or
Email your CV to nitin@secninjaz.com with details of your completed and ongoing agentic AI, multi-agent, evaluation or fine-tuning projects, your specific contributions, current CTC, expected CTC, and notice period or earliest joining date.
Email subject: Application — AI/ML Engineer — Agentic AI & Cybersecurity — [Your Name]
Click on Apply to know more.
