Updab - Software and data engineering consulting

Data engineering consulting

Try Updab's new AI tools!

SpecToMCP - a web app to generate Model Context Protocol servers from OpenAPI / Swagger specifications.

TestAgent - a RAG agent that lets you switch and compare models between GPT-4o Mini and a self-hostable Llama model.

About Updab

Updab AB, founded by William Saar (MSc) and incorporated in Sweden, offers AI-aware backend and data engineering services to international clients on a remote basis.

Experiences include data and backend engineering at scale for two consumer apps each serving 300M+ users, development of leading enterprise security and performance tools, and development and design of latency-sensitive financial trading tech.

Technologies include Apache Flink, Kafka Streams, Apache Beam, Google Dataflow, BigQuery, as well as older Hadoop-based frameworks, and programming languages Java, Rust, Scala, TypeScript, Python, Go, and SQL.

Backend and data engineering for Spotify and Candy Crush
Data engineering for appsec and cloudsec analytics and Rust development for Rapid7
Trading technology development for a Nasdaq-acquired marketplace platform powering exchanges such as Stock Exchange of Thailand and London Metal Exchange
Development for the Java performance-tuning suite that became Java Mission Control and shipped with Oracle's Java Development Kit
Helping San Francisco-based appsec startup tCell.io go from seed stage through Series-A to an acquisition by Rapid7
AI courses with certifications: Nvidia's Building RAG Agents with LLMs, DeepLearning.AI's Generative AI with Large Language Models, Stanford's Machine Learning
Patents in data analysis and software security

Projects

SpecToMCP - Generate Model Context Protocol (MCP) servers from Swagger / OpenAPI specifications https://spectomcp.updab.com

TestAgent - Python FastAPI and LangChain RAG agent with a Postgres vector store to compare AI models in a recruitment agency scenario https://testagent.updab.com

akhq-mcp-server - Open source MCP server to let AI assistants work alongside the user with Kafka monitoring tool AKHQ https://github.com/saarw/akhq-mcp-server

Open source project Flushout - a distributed data model based on event sourcing written in TypeScript
https://github.com/saarw/flushout

Mobile app ThinkCool - a now decommissioned consumer app developed with React Native and deployed to Apple and Google app stores https://thinkcool.app

GitBulletin - AI project in Rust that demonstrates using OpenAI APIs to generate newspaper-style front pages from Git commits, and Postgres vector stores with embedding models to search related activity.
https://gitbulletin.com

Writings and Talks

Slides from talk about Migrating batch-ETLs to Streaming Flink
https://www.slideshare.net/WilliamSaar/migrating-batch-etls-to-streaming-flink

tCell/Rapid7 blog article: Trusting Agents with Rust
https://blog.rapid7.com/2017/06/01/trusting-agents-with-rust/