Kaskada

What Kaskada does

Kaskada is an open-source, Python-native streaming engine that computes and serves temporal features for real-time AI and machine-learning applications. Its core product, the Kaskada Streaming Engine, is implemented in Rust using Apache Arrow for high-performance columnar computation and ships as an embedded Python library via PyPI under the Apache 2.0 license. The platform's primary abstraction, the Timestream, exposes a declarative API over event-time data and supports real-time aggregation, event detection, history replay, point-in-time joins, and windowing operations (tumbling, sliding, trailing, since), with incremental execution for fault-tolerance and late-data handling. Integration with Pandas and PyArrow is achieved through PyO3 bindings and zero-copy transfers. Architecture heritage traces to core contributors of Apache Beam, Google Cloud Dataflow, and Apache Cassandra.

The business is entirely open source with no commercial pricing tiers; distribution is product-led through PyPI and GitHub, while community engagement runs through GitHub Discussions, a Slack workspace, and scheduled Office Hours. Source data identifies zero disclosed paying customers, no certifications, no regulatory actions, and no named enterprise buyers. Kaskada's end users are AI/ML practitioners, data engineers, and generative-AI application builders who need to feed live event data into models and feature stores. Total disclosed equity raised prior to the DataStax acquisition was approximately $9.775 million across a 2018 seed round led by Voyager Capital and a 2020 Series A.

Kaskada is now a subsidiary of DataStax following a 2022 acquisition; DataStax open-sourced the core technology in March 2023 and the project was re-architected into a Python-native library in August 2023 (BeepGPT shipped as a showcase application that month). With 1-10 reported employees and no revenue model, the entity functions as a community-driven open-source project inside a larger strategic parent.

Name

Kaskada

Legal name

Kaskada

Website

https://kaskada.io

Company type

Private

Founded year

2020

Operating status

Operating

Headcount range

1–10 employees

Short description

Kaskada is an open-source, Python-native streaming engine that computes temporal features for real-time AI applications. Built in Rust on Apache Arrow and distributed via pip under Apache 2.0, it serves AI/ML practitioners and GenAI builders as a DataStax subsidiary.

Ownership category

akta.pro rank

Product category

Real-Time AI Data Streaming / Feature Engineering

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518), Computer Systems Design and Related Services (54151)

SIC

Services-Prepackaged Software (7372), Services-Computer Programming Services (7371)

akta.pro primary industry

Event Streaming & Pub/Sub Platforms (HDAEAIAD)

akta.pro secondary industries

Data & Analytics Platforms (Data Warehousing, Lakes, Streaming) (HDABAAAF), Webhooks, Eventing & Message Streaming (BPAMAOAE), Experiment Tracking, Metadata & Model Registry (HDAAABAC)

Keywords

• Real-time AI streaming
• Feature engineering platform
• Event processing engine
• Machine learning pipelines
• Time-series computation

Headquarters

HQ city

Seattle

HQ country

United States

HQ region

North America

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Operations, Marketing or Sales, Infrastructure

Revenue model

1. Open Source (Apache 2.0 License): Core Kaskada technology is open source under Apache 2.0 license, freely available for download and use. No direct revenue from the open source product itself.

Go-to-market motion2 records

Distribution channels2 records

Marketing channels6 records

Core offering

Kaskada is a next-generation streaming engine built in Rust using Apache Arrow, distributed as an embedded Python library that connects AI models to real-time and historical event data. Its core abstraction, the Timestream, provides a declarative Python-native query DSL for temporal feature engineering, including real-time aggregation, event detection, and history replay with point-in-time joins. The engine unifies batch and stream processing through a single API without requiring JVM infrastructure, enabling ML practitioners to compute features and trigger proactive AI behaviors from streaming event data.

Product overview

Kaskada is a real-time AI streaming engine built in Rust using Apache Arrow, released as an embedded Python library. The core product consists of the Kaskada Streaming Engine and its primary abstraction, the Timestream — a declarative API for time-centric computations over event-based data. Data sources (CsvString, JsonlString, Pandas, Parquet, PyDict) feed into Timestreams which support rich operations including aggregations (sum, count, mean, min, max, variance, stddev), time operations (shift_by, shift_to, seconds_since), windowing (Tumbling, Sliding, Trailing, Since), and execution modes (history, snapshot, live). The engine provides real-time aggregation, event detection, and history replay capabilities for ML feature engineering and real-time AI applications.

Differentiator

Problem solved

Functional benefit

Products and services

• Kaskada Streaming Engine A next-generation streaming engine built in Rust using Apache Arrow that connects AI models to real-time and historical data. Provides real-time aggregation, event detection, and history replay capabilities for time-centric computations over event-based data.
• Kaskada Python Library Embedded Python library providing the Kaskada compute engine, installable via 'pip install kaskada'. Includes Sources for data ingestion (CsvString, JsonlString, Pandas, Parquet, PyDict) and Destinations for output, enabling self-serve developer adoption without infrastructure provisioning.
• Timestream The core data structure of Kaskada representing how a value changes over time. Timestreams provide a declarative API over temporal context supporting aggregations (sum, count, mean, min, max, stddev, variance), time operations, arithmetic, string operations, comparisons, logical operations, records, collections, and grouping for time-centric feature engineering.

Segments2 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration6 records

AI capability4 records

Feature8 records

Scale indicators1 record

Recent moves6 records

Expansion highlights5 records

Direct peers

• Tecton: Tecton is the leading commercial feature platform for ML, offering feature engineering, storage, and serving for both real-time and batch features. It directly competes with Kaskada's feature computation and serving capabilities, though Tecton focuses on a managed, enterprise SaaS model rather than open source.
• Feast: Feast is an open-source feature store that manages feature computation, storage, and serving for ML models. It targets the same developer persona as Kaskada (data scientists and ML engineers) and overlaps significantly on feature engineering workflows, though Feast has a broader community and broader architectural footprint.
• Materialize: Materialize is a streaming SQL database that incrementally computes results from streaming data using PostgreSQL-compatible SQL. It competes directly with Kaskada's incremental query execution and real-time aggregation, with overlapping use cases in event-driven applications and analytics.
• Hopsworks: Hopsworks is an open-source ML platform with a strong feature store component. It directly competes with Kaskada in feature engineering for ML and shares the open-source go-to-market, though Hopsworks bundles a broader platform including model serving and notebooks.

Broad incumbents

• Apache Flink: Apache Flink is a mature, widely deployed open-source stream processing framework. While not ML-specific, it serves the same real-time event computation workloads Kaskada targets and is the dominant incumbent whose community and tooling Kaskada competes with for streaming mindshare.
• Apache Beam: Apache Beam is the unified programming model for batch and streaming data processing, on which Kaskada's founders previously worked. It competes for the same event-processing workloads but with a JVM-heavy SDK footprint that Kaskada deliberately avoids with its Python/Rust stack.
• Databricks Feature Store: Databricks' integrated Feature Store serves feature engineering and serving for ML models within the broader Databricks Lakehouse Platform. As a broad incumbent with massive enterprise distribution, it competes with Kaskada for organizational feature engineering workloads and can bundle features as part of a larger data platform deal.

Emerging players

• RisingWave: RisingWave is an open-source distributed SQL database for stream processing, built in Rust. It targets similar workloads to Kaskada (real-time aggregations, event processing) and shares the Rust + Arrow-style architectural philosophy, though it exposes a SQL interface rather than a Python DSL.
• Quix: Quix is a streaming data platform focused on ML and AI applications, using Python and Jupyter notebooks as the primary interface. It overlaps with Kaskada on real-time feature engineering and Python-native developer experience, though Quix emphasizes a managed cloud service.
• Pathway: Pathway is a Python framework for streaming data processing with built-in LLM integration. It targets the same real-time AI feature engineering use case as Kaskada and competes for Python-native data scientists building LLM applications on streaming data.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat4 records

Key risks6 records

Key highlights6 records

Customer concentration

• GitHubgithub.com/kaskada-ai/kaskada
• Xx.com/KaskadaInc
• Facebookfacebook.com/kaskadainc

Revenue estimate

Valuation estimate

Number of profiles

Profiles4 records

Funding overview

Funding rounds2 records

Investors4 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Kaskada