Architecture comparison

Lyft vs Uber

No shared foundational model — different bets.

Lyft

Mobility

Proximity

Off the LLM map (in-house non-LLM ML)

Models

—

Cloud · datastore

AWSMySQLDynamoDB

Compliance

SOC 2

Architecture

In-house pricing/ETA ML; off the LLM map at the core.

Uber

Mobility

Proximity

Off the LLM map (in-house non-LLM ML)

Models

—

Cloud · datastore

HybridSchemaless (MySQL-based Docstore)Docstore (on MySQL)MySQLCassandraRiakApache PinotHDFSHivePrestoApache SparkElasticsearchKafka (as streaming log/store)

Compliance

SOC 2 (inferred from large-scale hybrid cloud and enterprise focus)PCI-DSS (inferred from handling global payments and card data)GDPR (inferred from operating across EU with personal location data)

Architecture

Uber runs thousands of microservices on a hybrid cloud with multiple active data centers, using Docker on Mesos/Aurora, Kafka + Flink/Spark + Pinot/Presto/HDFS for real-time and batch data, and in-house layers like Michelangelo AI as a model gateway on top of this foundation.[11][16][3][9][6] Core operational data lives in Schemaless/Docstore on MySQL plus Cassandra/Riak, with extensive self-hosted open-source infrastructure (Kafka, Flink, Pinot, Spark, Presto, HDFS, Elasticsearch) augmented by proprietary platforms such as Michelangelo and Zero Growth Stack.[11][3][9][13][6]

← Full orbit map · Score your own stack →