I
Inworld AI Tech Stack
Realtime AI infrastructure for interactive agents and voice applications
Inworld AI builds infrastructure for realtime generative AI applications—voice models, agent runtimes, and multimodal inference at scale. The tech stack (PyTorch, vLLM, CUDA, Kubernetes, Ray, Terraform) reflects a systems-heavy engineering org optimizing for low-latency inference and multi-tenant serving. Active projects on model serving optimization, dynamic A/B experiments, and distributed inference scaling, paired with hiring across research and infrastructure roles, signal a company solving the hard problems of deploying interactive AI at concurrent scale rather than building consumer products.
Tech Stack 42 technologies
Core StackPython C++ PyTorch Rust Kubernetes Terraform ArgoCD Ansible Helm Go RAG Node.js Playwright C# Jenkins JavaScript TypeScript vLLM CUDA Ray NVIDIA GPU Terragrunt Kustomize GCP Azure Oracle Cloud Bash JavaScript/TypeScript Unreal Engine Unity+12 more
What Inworld AI Is Building
◆Challenges
- Handling thousands concurrent connections
- Sub-second multimodal inference at scale
- Zero to one
- Reliable agentic systems
- Realtime performance
- Shipping quickly
- Realtime online
- Expanding sales presence in north america
- Scaling solutions across industries
- Reliable serving of multimodal models
▲Active Projects
- Model serving optimization
- Api-based model services
- Dynamic a/b experiments
- System-wide billing
- Inference infrastructure scaling
- Realtime multimodal inference platform
- Realtime orchestration platform
- Distributed inference scaling
- Inference optimization for multimodal models
- Financial modeling and forecasting
Hiring Activity
15 roles · 10 in 30d
Department
Seniority
Company intelligence