AI Security Engineer

Dynanet Corporation•Remote, MD

7d•Remote

About The Position

Dynanet Corporation is seeking an AI Security Engineer to join their team. This role focuses on designing and implementing security architectures for AI workloads, including Large Language Models (LLMs) and AI agents, across various cloud environments. The engineer will be responsible for establishing defense-in-depth controls, implementing runtime guardrails, managing identity and access, ensuring secure software development lifecycles for AI, and operationalizing AI security governance and compliance frameworks. The position also involves AI red teaming, monitoring, incident response, and enabling stakeholders through training and guidance on responsible and secure AI practices.

Requirements

Azure (Azure OpenAI, AI Studio, AKS, Key Vault, Entra ID, Defender), Microsoft Purview, and M365 Copilot governance.
Experience with AWS (Bedrock, SageMaker, KMS) or GCP Vertex AI.
Hands-on guardrail implementation including content filters, safety classifiers, prompt injection defenses, jailbreak prevention, and tool whitelisting.
Securing RAG pipelines and vector databases (Cosmos DB + pgvector/FAISS, Pinecone, Weaviate).
OAuth/OIDC, SAML, SCIM, RBAC/ABAC; secrets management via Key Vault, Parameter Store, or Vault.
Encryption, tokenization, redaction, differential privacy basics, DLP-based PII/PHI detection.
Experience with data classification, retention, and lineage.
STRIDE threat modeling, secure coding, dependency scanning, secret scanning, SAST/DAST.
CI/CD for AI apps (GitHub Actions/Azure DevOps), IaC (Bicep/Terraform), policy-as-code (OPA/Conftest/Azure Policy).
Logging with Azure Monitor/Sentinel, tracing, metrics, and automated AI evaluation pipelines integrated with SIEM/SOAR.
Working knowledge of NIST AI RMF, ISO/IEC 42001, OWASP LLM Top 10, and public sector controls.
Experience documenting controls, audits, and risk assessments.
Proficiency in Python or TypeScript/Node.js.
Experience with agent/orchestration frameworks (LangChain, Semantic Kernel, Guidance, DSPy).
Strong written and verbal communication skills.
Highly organized with the ability to prioritize, balance, and effectively advance multiple competing priorities in a high-volume, fast-paced environment.
Ability to interact in a professional and collaborative manner with fellow Dynanet Teammates and the clients, and business partners that we work with.
Ability and desire to challenge and educate yourself to support and advance IT services delivery in the Federal agencies we serve.
Excellent judgment and creative problem-solving skills.
Respond to team member and client requests via email, MS teams, or other communication means during core business hours.
Active listening skills to understand clients' needs, and collaboration skills to work with other developers and designers.
Relevant degree in Computer Science, Engineering, Cybersecurity, or equivalent experience.

Nice To Haves

5–8+ years in application/cloud security with 2+ years in AI/ML or LLM security.
Experience enabling enterprise AI use cases or AI agents in regulated environments.
Familiarity with Microsoft Copilot for M365 governance and Microsoft Purview.
Experience with AI red teaming and building evaluation harnesses.
Exposure to privacy regulations (HIPAA, GLBA, GDPR/CCPA) and public-sector compliance.
Contributions to security frameworks or open-source guardrail tools
CISSP, CCSP, Azure Security Engineer (AZ-500), GIAC (GWEB/GWAPT/GXPN), OSCP.
Azure AI Engineer (AI-102), Azure Solutions Architect (AZ-305), AWS Security Specialty.
CISA, ISO 27001 Lead Implementer, Responsible AI certifications

Responsibilities

Design reference architectures for secure LLM/AI agent deployments across Azure, AWS, or hybrid environments.
Establish defense-in-depth controls for model endpoints, vector databases, prompt routing, tools/plugins, and orchestration layers.
Implement runtime guardrails including prompt injection defenses, output content filtering, PII detection/redaction, jailbreak prevention, and tool use restrictions.
Codify enterprise policies (acceptable use, data residency, retention, secrets handling) into enforceable controls through middleware, gateways, and policy engines.
Integrate Entra ID / Azure AD, OAuth/OIDC, and RBAC/ABAC models.
Apply data security measures including DLP, encryption, key management/HSM, tokenization, and fine-grained data access for RAG pipelines.
Embed threat modeling, secure coding, dependency scanning, secret scanning, and SAST/DAST into AI app pipelines.
Define AI-specific code review checklists for prompt templates, tool bindings, and agent plans.
Operationalize NIST AI RMF, ISO/IEC 27001 & 42001, SOC 2; align with FedRAMp, FISMA, NIST 800-53, and agency-specific controls.
Maintain model cards, data lineage, evaluation reports, and audit trails for AI decisions and tool calls.
Design adversarial tests for jailbreaks, prompt injections, data exfiltration attempts, and toxic outputs.
Build automated evaluation harnesses and metrics such as hallucination rates, sensitive content occurrence, and tool misuse rates.
Establish observability for AI systems including privacy-aware logging, policy hits, model drift detection, cost governance, and anomalies.
Define playbooks for AI incidents involving unsafe outputs, data leakage, compromised tools, or model endpoint abuse.
Partner with Product and Engineering teams to safely accelerate new AI use cases.
Provide training and guidance on responsible AI, secure agent design, and safe prompt engineering.