Poolside Tech Stack
LLM inference and agent infrastructure for enterprise software automation
Poolside operates a model-serving and software-agent platform built on PyTorch, Kubernetes, and NVIDIA GPU infrastructure. The tech stack—heavy on ML frameworks (JAX, vLLM, Triton), orchestration (Kubernetes, Terraform, ArgoCD), and observability (Datadog, Prometheus)—reflects a company focused on distributed LLM training, inference optimization, and deployment at scale. Active projects around multi-device inference, source-code generation, and deployment architecture evolution signal a shift from model-building toward production infrastructure; pain points around latency, throughput, and experimentation-pipeline velocity indicate the technical hurdles between research and reliable enterprise deployments.
Tech Stack 31 technologies
What Poolside Is Building
◆Challenges
- Secure deployment in hybrid environments
- Synthetic data generation
- Scaling deployment architecture
- Deployment tooling improvement
- Managing preventive maintenance programs
- Latency reduction
- Throughput improvement
- Slow experimentation pipeline
- Scalable large-scale data processing
- Improving dataset quality
▲Active Projects
- Distributed training and inference of llms
- End-to-end experimentation platform
- Deployment architecture evolution
- Secure infrastructure deployment
- Ai cooling deployments
- Massive-scale dataset delivery
- Multi-device inference of llms
- Source code generation inference
- Automation of provisioning and upgrades
- Major architecture changes
