Autonomous Defense Research Intern

You will design and prototype autonomous agents that learn to defend enterprise infrastructure against evolving threats. Working at the intersection of AI safety, cybersecurity, and reinforcement learning, your research will ship into Cehpoint's defense products and peer-reviewed outputs — directly shaping how autonomous systems operate in adversarial environments.

• Prototype and train reinforcement-learning or multi-agent architectures for autonomous threat detection, triage, and response simulation
• Conduct red-team/blue-team exercises to stress-test autonomous defense agents against realistic attack scenarios (MITRE ATT&CK-aligned)
• Research and integrate defensive adversarial-ML techniques (prompt injection mitigation, data-poisoning resistance, model watermarking)
• Build and maintain evaluation harnesses: synthetic CTF-style environments, attack simulation pipelines, and benchmark dashboards
• Publish concise research briefs and internal tech-notes; contribute to open-source tooling and patent disclosures where applicable
• Collaborate with senior engineers to transition research prototypes into production-grade inference pipelines (Docker, Kubernetes, CI/CD)

• Foundations of reinforcement learning (MDPs, policy gradient, Q-learning) and their application to sequential decision-making under uncertainty
• Core cybersecurity concepts: network traffic analysis, common attack vectors, incident-response lifecycle, and security-ops tooling
• Adversarial machine learning: attack surface of ML models, defense mechanisms (adversarial training, input sanitization, formal robustness bounds)

• Python — production-quality code with pytest, logging, and packaging; PyTorch or JAX for model development
• MLOps & containerization — Docker, MLflow or Weights & Biases, basic Kubernetes for deployment
• SIEM / log-analysis familiarity — elastic-stack (Elasticsearch, Kibana), Splunk queries, or equivalent
• Git, Linux CLI, and infrastructure-as-code basics (Terraform or Pulumi for spinning up sandbox environments)

• A track record of building autonomous agents or multi-agent simulations outside coursework (open-source contributions, Kaggle, HackMIT, etc.)
• Genuine fascination with the intersection of AI safety and offensive security — you read both NeurIPS papers and offensive-sec write-ups
• A habit of shipping research prototypes end-to-end, not just notebooks — you care about latency, observability, and real-world deployability

• Competitive stipend (₹12L–₹20L/year) with performance-based acceleration and a conversion path to full-time researcher/engineer
• Fully remote setup with a ₹15,000 home-office stipend and access to enterprise-grade GPU credits (A100/H100 pool)
• Direct mentorship from PhD-led researchers and veterans of top-tier security teams (C-DAC, DRDO, NCRB, and Fortune-500 CISO offices)
• Publication and patent support — co-author opportunities on peer-reviewed venues and internal IP filings
• Annual research retreat in India with full travel and accommodation covered; access to international conference sponsorship