Senior AI Engineer (Agentic Systems & Inference) - Onsite

Full Description

As a Senior AI Engineer, you will be the primary architect of Cognna’s autonomous agent reasoning engine and the high-scale inference infrastructure that powers it.

You are responsible for building production-grade reasoning systems that proactively plan, use tools, and collaborate.

You will own the full lifecycle of our specialized security models, from domain-specific fine-tuning to architectingdistributed, high-throughput inference services that serve as Security-core intelligence in our platform.

Key Responsibilities

Agentic Architecture & Multi-Agent Coordination

• Autonomous Orchestration: Design stateful, multi-agent systems using frameworks like Google ADK.

• Protocol-First Integration: Architect and scale MCP servers and A2A interfaces, ensuring a decoupled and extensible agent ecosystem.

• Cognitive Optimization: Develop lean, high-reasoning microservices for deep reasoning, optimizing context token usage to maintain high planning accuracy with minimal latency.

2. Model Adaptation & Performance Engineering

• Specialized Fine-Tuning: Lead the architectural strategy for fine-tuning open-source and proprietary models on massive cybersecurity-specific telemetry.

• Advanced Training Regimes: Implement Quantization-Aware Training (QAT) and manage Adapter-based architectures to enable the dynamic loading of task-specific specialists without the overhead of full-model swaps.

• Evaluation Frameworks: Engineer rigorous, automated evaluation harnesses (including Human annotations and AI-judge patterns) to measure agent groundedness and resilience against the Security Engineer’s adversarial attack trees.

3. Production Inference & MLOps at Scale

• Distributed Inference Systems: Architect and maintain high-concurrency inference services using engines like vLLM, TGI, or TensorRT-LLM.

• Infrastructure Orchestration: Own the GPU/TPU resource management strategy.

• Observability & Debugging: Implement deep-trace observability for non-deterministic agentic workflows, providing the visibility needed to debug complex multi-step reasoning failures in production.

4. Advanced RAG & Semantic Intelligence

• Hybrid Retrieval Architectures: Design and optimize RAG pipelines involving graph-like data structures, agent-based knowledge retrieval and semantic searches.

• Memory Management: Architect episodic and persistent memory systems for agents, allowing for long-running security investigations that persist context across sessions.

• Experience & Education: 5-7+ years in AI/ML Engineering or Backend Systems.

Must have contributed to large-scale AI/ML inference service in production.

M.S. or B.S. in Computer Science/AI is preferred.

• Inference Orchestration: KV-cache management, quantization formats like AWQ/FP8, and distributed serving across multi-node GPU clusters).

• Agentic Development: Expert in building autonomous systems using Google ADK/Langgraph/Langchain and experienced with AI Observervability frameworks like LangSmith or Langfuse. Hands-on experience building AI applications with MCP and A2A protocols.

• Cloud AI Native: Proficiency in Google Cloud (Vertex AI), including custom training pipelines, high-performance prediction endpoints, and the broader MLOps suite.

• Programming: Python and experience with high-performance backends (Go/C++) for inference optimization. You are comfortable working in a Kubernetes-native environment.

• CI/CD: You are comfortable working in a Kubernetes-native environment.

💰 Competitive Package – Salary + equity options + performance incentives

🧘 Flexible & Remote – Work from anywhere with an outcomes-first culture

🤝 Team of Experts – Work with designers, engineers, and security pros solving real-world problems

🚀 Growth-Focused – Your ideas ship, your voice counts, your growth matters

🌍 Global Impact – Build products that protect critical systems and data