Senior Data Engineer

Trust In SODA; Greater London; Hybrid; GBP 74,000 - 83,000; Part time

Job summary

Trust In SODA is seeking a data-focused engineer to design data models and build production-grade Python pipelines. You will work with inputs and outcomes, shaping schemas and storage layouts, then implement end-to-end data flows with light review.

You will collaborate with the Head of Data and Lead Data Scientist/Engineer, ensuring production quality and coherent architecture in a startup-like environment. 3-4 months contract with on-site/remote balance.

Qualifications

• Strong production Python experience in data projects.
• Experience designing data models and schemas.
• Ability to define schemas and storage layouts.
• Able to work with incomplete requirements and deliver pipelines.
• Proven track record shipping data/ML systems in production.
• Experience in startup or small product teams.

Responsibilities

• Turn client data (APIs, CSVs, S3) into reliable Python pipelines.
• Specify schemas and storage layout (Parquet on S3, layered approach).
• Orchestrate jobs in Python; Dagster used; other solid Python jobs accepted.
• Work on AWS; Terraform/EKS knowledge not required.
• Use AI coding agents heavily and defend architecture and data model.

Skills

Python

Data modeling

Production pipelines

Ambiguity handling

AI-assisted development

Startup experience

Data engineering in production

Tools

Dagster

Airflow

Prefect

Parquet

S3

Terraform

Job description

Location: London Bridge - In office 2-3 days per week

Start Date: ASAP

Duration: 3-4 months with extension

Daily Rate: £400 - £450 per day outside IR35

Summary

You will sit with the Head of Data and the Lead Data Scientist/Engineer. You will be pointed at inputs and expected outcomes, then expected to design and build the path between them - including the data model - with light review.

The work still must be grounded: clear schemas, sensible storage layout, production-quality Python. It is not cowboy scripts, and it is not waiting for a backlog of tickets.

This is a bad fit if you mainly plug enterprise components together, wait for JIRA epics, or treat AI coding tools as a novelty. This is a good fit if you have built data/ML systems in a startup or small product team, you use Cursor/Copilot (or equivalent) as a normal part of shipping, and you can own a problem from messy source files to a running pipeline without being sequenced

Requirements

• Strong production Python
• Evidence of designing data models and schemas, not only consuming them
• Comfort operating with incomplete requirements: inputs and outcomes, then you fill in the middle
• Can take messy inputs and an expected outcome, then design schema + build the pipeline with light review
• Evidence of designing a production pipeline from messy source data, not just orchestrator config
• AI-assisted development as a default way of working, not a talking point
• Using AI coding tools (Cursor, Copilot or equivalent as a normal way of shipping
• 4+ years shipping data or applied ML systems in production
• Previous experience in a start-up or a small product team

• Dagster, or Airflow, or Prefect in production
• Data lakes / Parquet / S3
• Terraform or general cloud familiarity (infra is owned by another team)
• RAG, embeddings, or other LLM-adjacent pipelines
• Startup or small-team product delivery

• Turn client data (APIs, CSVs, S3, messy operational exports) into reliable Python pipelines.
• Specify schemas and storage layout (Parquet on S3, layered / medallion-style) so the next person can extend the work.
• Orchestrate jobs in Python. We use Dagster; Airflow, Prefect, or well-structured Python jobs are fine.
• Work on AWS. You do not need to own Terraform, EKS, or networking.
• Use AI coding agents heavily, then stand behind the architecture and the data model.
• Shape approach with the rest of the data team: enough design to stay coherent, then execute at speed

• Assembling warehouse / lakehouse platforms (Spark, Informatica, "I wired Airflow to the lake")
• Writing TDDs and JIRA epics rather than shipping code
• Large bank / SI / programme delivery with little product ownership
• ML research / model-training CVs with no real-world data engineering