Full-time · Remote (India)

AI Behavior & Alignment Architect

Cehpoint is hiring for this role: full-time, Remote (India), ₹13,00,000 – ₹22,00,000 / year. Formulate formal behavioral boundaries, automated red-teaming harnesses, constitutional AI principles, and steerability systems for mission-critical client deployments.

About the role

As autonomous agents gain access to enterprise databases, shell environments, and transactional systems, behavioral failure is catastrophic. You will engineer mathematical boundaries and runtime monitors that guarantee agent actions remain compliant with corporate governance and legal mandates.

We are seeking an experienced AI Behavior Architect to define the behavioral rules of engagement for our enterprise AI systems. You will lead adversarial red-teaming, constitutional policy enforcement, and multi-agent governance protocols.

What you will do

  • Author formal Constitutional AI principles and runtime guardrails for autonomous tool-execution loops.
  • Design automated adversarial attack simulations (jailbreaks, prompt injections, privilege escalations, context contamination).
  • Build real-time behavioral interceptors that inspect tool arguments before OS/API execution occurs.
  • Publish formal System Cards and Risk Assurance Dossiers for enterprise client boards and regulatory compliance audits.
  • Establish automated regression testing suites that prevent behavioral degradation during model upgrades.

What you should know

  • In-depth knowledge of LLM vulnerability taxonomies: OWASP Top 10 for LLMs, NIST AI RMF, and MITRE ATLAS.
  • Theoretical understanding of value alignment, reward hacking, specification gaming, and goal misgeneralization.
  • Familiarity with regulatory frameworks including India's DPDP Act, EU AI Act, and enterprise ISO standards.

What you should be able to do

  • Proficiency in Python for building automated security test harnesses and fuzzing pipelines.
  • Hands-on experience deploying guardrail toolkits (NVIDIA NeMo Guardrails, Llama Guard, custom AST parsers).
  • Strong technical writing skills for authoring institutional safety reports.

How you work

  • Relentless adversarial mindset that anticipates edge cases before they manifest in production.
  • Commitment to practical, high-throughput security that does not needlessly strangle agent utility.
  • Continuous research into newly published alignment failures and prompt injection vectors.

Skills we check

  • Constitutional AI
  • Red Teaming Automation
  • Adversarial Prompting
  • Steerability Bounds
  • Safety Benchmarking
  • Python
  • LLM Guardrails
  • System Card Design

What you get

  • Annual compensation ₹13,00,000–₹22,00,000 + performance bonus.
  • High visibility role presenting directly to enterprise CISOs and engineering executives.
  • Fully remote setup with dedicated research compute allocation.
  • Generous annual training budget for advanced security and AI certifications.

How we select for this role

We read every application ourselves. Here is what happens after you apply.

  1. Apply with a short form

    Paste a link to your CV and answer five questions in your own words. No sign-in and no file upload. About fifteen minutes.

  2. Integrity check

    When you press Check my answers, an automated check confirms the answers were typed by you. You see the result straight away, before you submit.

  3. Choose your project date

    If no cheating is detected you are shortlisted, and we email you a link to choose the day your project starts.

  4. 5-day practical project

    On that day our team contacts you with the brief: real work, 5 working days, reviewed by an engineer. For this role: Construct an automated adversarial fuzzing suite that tests autonomous agents against jailbreaks, unauthorized system execution, and data exfiltration vectors, generating automated compliance reports (strictly 5 working days).

  5. Offer

    If your work solves the problem well, you receive your offer letter by email and in the portal.