Simon Thelin

Multimodal ML Platform · Data Engineer · AI Creator

Building the data infrastructure that trains the next generation of multimodal AI — at Synthesia and beyond. Over a decade in data and ML space.

Data Architecture End To End Lakehouse Architecture DataOps RAG & MCP

~/datapains.io git:(main) ❯ █

Industry experience

Infra
costs cut

Successfully reducing infrastructure cost across Data & AI projects

Query latency reduction

Global
R&D

Serving researchers across countries · Synthesia

About

End-to-end ML Platform
& Data Engineer.
AI creator.
DataOps fundamentalist.

10+ years building scalable, governed data systems across AWS and GCP — Lakehouse architecture, DataOps pipelines, and production-grade ML infrastructure. Currently Tech Lead at Synthesia, serving R&D researchers across multiple countries as part of a global AI video platform.

DataPains is where I share what I've learned — practical, no-fluff content on data engineering, AI tooling, and the platforms that actually matter in production. Conference speaker at Big Data London and DataNova 2023. Featured on the Data Team Success podcast.

Python Terraform AWS GCP dbt Airbyte Trino PySpark Kafka K8s Delta Lake Airflow ArgoCD Docker LLMs AI Video

Content

DataPains on YouTube

@DataPains

ML PLATFORM · DATA ENGINEERING · AI · TOOLS THAT MATTER IN PRODUCTION

● — subscribers ● — videos ● — total views

Master Apache Spark on Kubernetes and Beyond!

DATAPAINS · YOUTUBE

Unlocking the Power of ArgoCD: A Game Changer for Kubernetes Users

DATAPAINS · YOUTUBE

Think Twice: Drawbacks of Apache Airflow Revealed

DATAPAINS · YOUTUBE

View all videos ↗

Blog

Latest Writing

MEDIUM

F3 Fixes Tables, But Does It Solve Unifying Vectors?

A look at whether F3's table format improvements actually address the deeper challenge of unifying vector and structured data in modern AI stacks.

Read on Medium →

MEDIUM

Data Architecture: Managing Inactive Data

How to think about cold, inactive data in a Lakehouse — lifecycle policies, tiering strategies, and the cost implications of keeping everything hot.

Read on Medium →

MEDIUM

How to Think About Data Modelling and Semantic Layers

A mental model for structuring data transformations and semantic layers — where dbt fits, where it doesn't, and how to draw the right boundaries.

Read on Medium →

All posts on Medium →

Experience

Career

Jul 2024 — Present

Tech Lead — DataOps within ML Platform

Synthesia · London

Owns DataOps strategy and execution enabling multimodal ML training (audio + video) across AWS and GCP. Delivered self-service ML training framework with full data lineage and provenance tracking. Architected Lakehouse platform (Delta + Trino) with federated query across distributed data domains. Migrated WEKA to S3, reducing infrastructure spend. Led human annotation data evaluation platform — A/B and MUSHRA evaluation frameworks with structured feedback loop patterns for continuous inference quality assessment. Exploring RAG/TAG (Tool Augmented Generation) for MCP-driven dataset and evaluation workflows.

AWS GCP Delta Lake Trino MLflow Airflow Kubeflow Argo Workflows ArgoCD Spark PyTorch Terraform K8s Python

May 2022 — Jul 2024

Lead Data Engineer

7bridges · London

50% reduction in compute and storage costs. 98% query latency reduction via Trino-based Lakehouse architecture on GCP. Championed end-to-end DataOps platform build across batch and streaming. Conference speaker at Big Data London and featured in Starburst's DataNova 2023 success story.

GCP Trino Delta Lake dbt Airbyte Spark Kafka Airflow K8s

Dec 2021 — May 2022

Lead Data Engineer (Consultant)

Kainos · London

Contracted to provide data platform architecture guidance and hands-on technical leadership to enterprise clients.

AWS Python Terraform

Aug 2020 — Dec 2021

Data Engineer

IMG Arena · London

Built real-time sports data pipelines processing millions of events across streaming and batch. Designed Lakehouse architecture on Apache Druid. PySpark, Kafka, and AWS-native services powering live sports analytics products.

PySpark Kafka AWS Apache Druid Airflow

Dec 2019 — Jun 2020

Data Engineer

GameAnalytics · London

Improved PySpark/Hive data pipelines into BigQuery. Built the company's data science platform. Introduced Apache Airflow from scratch and established data engineering best practices.

PySpark BigQuery GCP Airflow Python

Speaking & Press

On stage & on the record

CONFERENCE TALK ↗

Big Data London

Live conference talk on Lakehouse architecture and DataOps in production — one of Europe's largest data engineering events.

Watch on YouTube →

CONFERENCE · STARBURST ↗

DataNova 2023

Building a data analytics platform with a Lakehouse at 7bridges — featured by Starburst as a DataNova success story.

Read the story →

PODCAST ↗

Data Team Success

In conversation with Ross Webb — on building high-performing data teams, ML Platform strategy, DataOps culture, and the realities of end-to-end data engineering at scale.

Listen to the episode →

PUBLISHED RESEARCH ↗

Indirect Tire Monitoring System — Machine Learning Approach

Published in IOP Conference Series: Materials Science and Engineering. Early application of machine learning to real-world sensor data for indirect tyre pressure monitoring.

View on Google Scholar →

Contact