Principal ML Architect – Security AI & Advanced Model Systems

Proofpoint•Sunnyvale, CA

2d•Hybrid

About The Position

We are seeking a Principal ML Architect to lead the design and development of next-generation AI systems for cybersecurity, leveraging state-of-the-art LLMs/SLMs and advanced machine learning techniques. This role requires deep expertise in model architecture, training, fine-tuning, and distillation, combined with a strong understanding of security domains such as threat detection, anomaly detection, data protection, and AI safety. You will drive the development of intelligent, security-focused AI systems and agents capable of operating at scale across high-volume, adversarial, and sensitive environments, while ensuring robustness, explainability, and compliance.

Requirements

10+ years in ML/AI systems, with significant focus on deep learning and production ML
Proven experience in: Building or scaling LLMs/SLMs or advanced ML systems, Applying ML/AI in security, fraud, risk, or adversarial domains
Track record of delivering production-grade AI systems at scale
Deep understanding of: Transformer architectures and modern LLM techniques, Retrieval-augmented generation (RAG) and hybrid AI systems, Model training dynamics, scaling laws, and optimization
Hands-on experience with: Training, fine-tuning, and distilling models, Efficient inference (quantization, pruning, batching), Distributed training frameworks (PyTorch, DeepSpeed, FSDP, etc.)
Strong understanding of one or more: Security telemetry (logs, network traffic, endpoint data), Threat detection and anomaly detection systems, Identity, access, and data protection systems
Familiarity with security tooling ecosystems (SIEM, EDR, CASB, etc.)
Experience designing high-throughput, low-latency ML systems
Strong programming skills in Python, with production experience
Understanding of data pipelines, feature engineering, and MLOps practices

Nice To Haves

Experience building AI systems for SaaS security or GenAI security platforms
Familiarity with multi-agent systems for security automation
Experience with synthetic data generation for security use cases
Contributions to AI/ML research, open-source, or security tooling
Background in AI safety, adversarial ML, or model interpretability

Responsibilities

Design and lead architecture for AI-driven security platforms, including: Threat detection and behavioral analytics, Data loss prevention (DLP) and insider risk detection, AI usage monitoring and policy enforcement (GenAI security)
Build systems that process high-volume, high-velocity security telemetry in real time
Lead development of state-of-the-art SLMs/LLMs tailored for security use cases: Log analysis, alert triage, threat intelligence, policy reasoning
Drive experimentation with modern architectures (Transformers, MoE, retrieval-augmented systems, hybrid models)
Balance trade-offs between model accuracy, latency, interpretability, and cost
Architect pipelines for: Domain adaptation and instruction tuning on security-specific datasets, Model distillation and compression for efficient deployment in enterprise environments
Design and execute experiments for: Alignment (RLHF/RLAIF) in security-sensitive contexts, Red-teaming and adversarial robustness of models
Design and oversee AI agents that: Automate security operations (SOC workflows, triage, investigation), Integrate with enterprise tools (SIEM, EDR, SaaS platforms)
Define architectures for tool use, reasoning, memory, and policy-aware decision making
Establish rigorous evaluation frameworks for: Detection accuracy, false positives/negatives, Model robustness under adversarial conditions, Safety, hallucination, and misuse risks
Lead deep experimentation cycles to continuously improve model performance and reliability
Guide deployment of models into enterprise-scale, real-time environments
Optimize inference systems for low latency, high throughput, and cost efficiency
Collaborate with platform teams on ML infrastructure, data pipelines, and observability
Ensure models and systems meet enterprise security standards (SOC2, ISO, GDPR, etc.)
Establish best practices for: Secure model development and deployment, Data privacy and protection in training pipelines, Responsible AI and model safety in adversarial environments

Benefits

Competitive compensation
Comprehensive benefits
Career success on your terms
Flexible work environment
Annual wellness and community outreach days
Always on recognition for your contributions
Global collaboration and networking opportunities
Flexible time off
Comprehensive well-being program with two paid Wellbeing Days and two paid Volunteer Days per year
Three-week Work from Anywhere option