Backend AI Engineer

This is a remote position.

Position Overview:

We are seeking an elite Lead Full Stack AI Engineer to design and build the operator control plane for our enterprise AgenticOps platform. In this role, you will connect real-time web interfaces with asynchronous multi-agent backends—turning complex agent reasoning, dynamic network topologies, and automated remediation runs into intuitive, transparent interfaces for IT and SOC operators. You will bridge frontend performance with modern AI frameworks, NVIDIA/Cisco integrations, and open-source tooling.

Core Responsibilities:

• Operator Cockpit & HITL Control: Build the primary AgenticOps console with real-time telemetry streaming, interactive multi-agent reasoning traces, execution dry-runs, and human-in-the-loop (HITL) approval/override controls.
  
• Dynamic Topology & DAG Visualization: Develop high-performance WebGL/canvas-based visual interfaces to render large-scale enterprise network maps, asset dependencies, and live multi-agent execution graphs without UI lag.
  
• Low-Latency Streaming Pipelines: Engineer bidirectional streaming protocols (WebSockets, SSE, gRPC-Web) to handle real-time agent thought streams, tool invocation payloads, and live terminal outputs.
  
• BFF & Edge Middleware: Build performance backend-for-frontend (BFF) layers in Go, Python (FastAPI), or TypeScript, implementing session state persistence, token caching, SAML/OIDC authentication, and fine-grained RBAC.
  
• AI Tooling & Ecosystem Integration: Connect the UI layer to agent runtimes, open-source automation tools (OpenClaw, Browser-Use), and model endpoints powered by NVIDIA NIMs, vLLM, and local SLMs.
  
• AI-Augmented SDLC: Utilize AI-assisted development workflows, automated component scaffolding, synthetic UI testing, and continuous eval pipelines to accelerate delivery.
  

Required Qualifications:

• Experience & Track Record: 8+ years building enterprise-grade full-stack web applications, with 3+ years focused on AI/LLM-native interfaces, observability platforms (APM/SIEM), or network operations tooling.
  
• Frontend Architecture: Deep mastery of TypeScript, React / Next.js, Tailwind CSS, state management (Zustand, TanStack Query), and rendering optimization for real-time streaming data.
  
• Graph & Canvas Visualization: Production expertise with canvas/SVG/WebGL visualization libraries (React Flow, Cytoscape.js, PixiJS, D3.js) for large-scale hierarchical graphs and multi-step agent reasoning trees.
  
• AI & Agentic Ecosystem: Hands-on experience with modern agent frameworks (LangChain, LangGraph, Vercel AI SDK), open-source agentic tools (OpenClaw, Browser-Use), and consuming endpoints from NVIDIA NIMs, vLLM, or Ollama.
  
• Streaming & Real-Time Protocols: Production fluency with SSE, WebSockets, and chunked streaming, managing partial token rendering and optimistic UI states under network latency.
  
• UX for Autonomous Systems: Strong design intuition for non-deterministic software: progressive disclosure of reasoning, confidence/risk scoring, before-and-after diff previews, and safety-critical confirmation workflows.
  

Mandatory:

Experience

• 8+ years building enterprise-grade full-stack web applications
  
• 3+ years on AI/LLM-native interfaces, observability platforms (APM/SIEM), or network operations tooling
  

Frontend

• TypeScript, React / Next.js, Tailwind CSS
  
• State management: Zustand, TanStack Query
  
• Rendering optimization for real-time streaming data
  

Graph & Visualization

• Production experience with at least one of React Flow, Cytoscape.js, PixiJS or D3.js (canvas/SVG/WebGL)
  
• Rendering large hierarchical graphs and multi-step agent reasoning trees
  

Real-Time Streaming

• SSE, WebSockets and chunked streaming
  
• Partial token rendering and optimistic UI under network latency
  

Backend / BFF

• Backend-for-frontend layers in Go, Python (FastAPI) or TypeScript
  
• SAML/OIDC authentication, RBAC, session state and token caching
  

AI & Agentic Ecosystem

• Agent frameworks: LangChain, LangGraph or Vercel AI SDK
  
• Consuming model endpoints from NVIDIA NIMs, vLLM or Ollama
  

UX for Autonomous Systems

• Human-in-the-loop (HITL) approval/override workflows
  
• Progressive disclosure of reasoning, confidence/risk scoring, diff previews and safety-critical confirmations
  

Good to Have (Optional)

• gRPC-Web
  
• OpenClaw, Browser-Use
  
• Local SLM deployment
  
• NVIDIA/Cisco platform integrations