P
Pathway is building an alternative to Transformers paired with a data processing engine, positioning itself as a post-transformer frontier model designed to solve memory constraints in AI systems. The tech stack—PyTorch, JAX, TensorFlow, Hugging Face, plus enterprise infrastructure (Kubernetes, Airflow, SageMaker, Vertex AI)—reflects a company scaling both model R&D and production ML systems. Pain points around GPU scaling, dataset management at terabyte scale, and model performance monitoring indicate they're solving hard infrastructure problems rather than wrapping existing APIs.
Tech Stack 28 technologies
Core StackLinux Docker Kubernetes GitHub Actions Jenkins GitLab CI/CD AWS SageMaker Vertex AI Grafana CloudWatch Prometheus Loki Terraform CloudFormation MLflow Kubeflow Apache Airflow PyTorch TensorFlow RAG Hugging Face Slurm GCP Azure Metaflow Jax CI/CD
AdoptingRAG
What Pathway Is Building
◆Challenges
- Managing terabyte-large datasets
- Cost performance optimization
- Scaling gpu clusters
- Scaling infrastructure to growing team
- Automating ml pipelines
- Monitoring model performance drift
- Customer adoption
- Large scale training
- Decreasing time-to-deployment
- Technical success
▲Active Projects
- Ml pipeline automation
- Ml training and inference infrastructure optimization
- Ml ci/cd implementation
- Evaluation infrastructure for post-transformer models
- Dataset pre-processing for llm accuracy
- Benchmarking process for ai model evaluation
- Implement ml-centric ci/cd
- Developer relations strategy
- Develop reusable solution patterns
- Model leaderboard
Hiring Activity
30 roles · 7 in 30d
Department
Seniority
Company intelligence
