Kumo Tech Stack

AI platform for training models directly on relational data

Software Development Mountain View, California 51–200 employees Founded 2021 Privately Held

Kumo builds a platform that trains machine learning models on structured data without manual feature engineering. The tech stack reveals a mature ML infrastructure play: PyTorch and TensorFlow for modeling, Kubernetes and cloud orchestration (AWS, Azure, GCP) for scaling, and Prometheus/Grafana for observability. Active projects span training pipelines, distributed inference systems, and multi-tenant infrastructure — all pointing toward enterprise ML deployment as the core use case. Hiring is engineering-led (6 of 13 roles) with a senior-skewed team, though velocity is decelerating.

Tech Stack 19 technologies

Core StackSalesforce Python Java C++ AWS PyTorch TensorFlow Kubernetes Prometheus Grafana OpenTelemetry Terraform Pulumi ArgoCD Apache Spark Azure GCP Flux Argo Workflows

What Kumo Is Building

◆Challenges

Leveraging fraction of data for machine learning
Slow ml iterations
Complex feature engineering
Disappointing ml results
Scaling multi-tenant ai platform
Cost optimization across clouds
Reducing developer toil
Underutilized data storage
Scaling predictive ai deployments
Enterprise ml deployment challenges

▲Active Projects

Training pipelines
Ai-powered user-facing products on rfm
Ai agents for data scientists on relational data
Integration of rfm with enterprise data systems
Distributed training and inference systems
Integration between data warehouses and ml engines
Cloud-native infrastructure
Real-time inference clusters
Ci/cd for large ml workloads
Multi-tenant infrastructure for ai workloads

Hiring Activity

Decelerating15 roles · 3 in 30d

Department

Engineering

Data

Sales

Other

Seniority

Senior

Mid

Lead

Company intelligence

Find more companies like Kumo by tech stack, pain points and active projects

Get started free

About Kumo

Kumo is a machine learning platform company founded in 2021 and based in Mountain View, California. The product enables data teams to build AI models on relational data (SQL databases, data warehouses) without manual feature engineering — a technical constraint that historically slowed ML adoption in enterprises. The company markets use cases including recommendations, fraud detection, risk scoring, entity resolution, and retrieval-augmented generation. Engineering and data teams form the largest hiring cohorts; the company is currently at 51–200 employees and hiring primarily in the United States.

HeadquartersMountain View, California

Company Size51–200 employees

Founded2021

Hiring MarketsUnited States

Frequently Asked Questions

What tech stack does Kumo use?

Kumo uses PyTorch and TensorFlow for model training, Python and Java for backend services, Kubernetes for orchestration, AWS/Azure/GCP for cloud infrastructure, and Prometheus/Grafana for monitoring. Terraform and Pulumi handle infrastructure-as-code.

What is Kumo working on?

Active projects include training pipelines, distributed training and inference systems, real-time inference clusters, multi-tenant infrastructure, and integration between data warehouses and ML engines. The company is also building AI agents for data scientists and CI/CD tooling for large ML workloads.