Skip to main content

AI & Data Engineering

Data pipelines and end-to-end AI/ML — from observation to insight, edge to ground.

End-to-End
Data Flow
ML Lifecycle
Demonstrated
Since 2016
Operational

We build software that follows data from observation to insight. Calibration pipelines that turn raw telemetry into science products. Training systems that turn labeled datasets into deployable models. Analytics that turn archives into discoveries.

Our work spans the full data lifecycle: processing, learning, and exploitation — on the ground, at the edge, or in orbit.

We don't build mission control systems. We build what happens to the data after it arrives.

Featured Talk

AI in Space – Mission Resilience and Autonomy

Presentation exploring onboard AI deployment, real-time orbital data processing, and autonomous decision-making in resource-constrained space environments.

on {ai}r | AI for Space — organized by Ateleris Swiss AI Weeks, September 2025

Our Approach

What We Build

Data Pipelines

Telemetry parsing, calibration, Level 0–2 processing, batch orchestration

AI/ML Systems

Use case evaluation, dataset curation, model training, edge deployment

Archives & Access

Long-term storage, query interfaces, data distribution systems

Where It Runs

Ground Systems

HPC clusters, cloud platforms, on-premise deployment

Edge Systems

Onboard processing, embedded inference, resource-constrained platforms

We select the architecture that fits your mission and data volume.

Capabilities

Data Pipelines

Telemetry parsing, bulk calibration, Level 0 through Level 2 processing, batch orchestration, quality assurance.

AI/ML Systems

Use case evaluation, dataset curation and labeling, training infrastructure, quantization-aware optimization, edge and ground deployment.

Archives & Access

Long-term storage, query interfaces, data distribution, community access portals.

Edge Computing

Onboard processing, embedded inference, bandwidth-efficient operations, autonomous alerting.

End-to-End ML Lifecycle

1

Scope

Define & Assess

2

Curate

Collect, Label & Split

3

Engineer

Features & Pipeline

4

Train

Fit & Optimize

5

Validate

Test & Evaluate

6

Deploy

Serve & Monitor

From use case to production — we own the full lifecycle, whether deploying onboard a satellite, on a drone, or in a ground-based environment.

Technologies

Languages

  • Python
  • Scala
  • C / C++

Data & Processing

  • PostgreSQL
  • Time-series DBs
  • Apache Spark

ML Frameworks

  • PyTorch
  • TensorFlow / TFLite
  • YOLOX
  • Quantization tools

Labeling & Auto-Annotation

  • CVAT
  • Segment Anything (SAM)

MLOps & Data Versioning

  • DagsHub
  • DVC
  • MLflow
  • Airflow

Edge Platforms

  • Xilinx Zynq / UltraScale+
  • Vitis AI / DPU
  • Buildroot / Embedded Linux

Infrastructure

  • HPC clusters
  • Kubernetes
  • Cloud platforms

Standards

  • CCSDS
  • FITS
  • PUS-C

Project Heritage

STIX Ground Pipeline

Operational

Scientific data pipeline for the STIX X-ray telescope on ESA's Solar Orbiter. Processing, calibration, archiving, and distribution to the science community. From raw telemetry (L0) to science-ready products (L2).

ESA PRODEX FHNW (Prime) 2016 – ongoing

Onboard Object Detection

Demonstrated

Real-time AI inference at 10.5 Mpx/s on resource-constrained platforms. YOLOX optimized for Xilinx UltraScale+ at 26W power budget.

armasuisse S+T Prime 2022 – 2024
View product

Image Preprocessing Pipeline

Completed

TensorFlow-based onboard image preprocessing for embedded satellite payloads. Demosaicing, calibration, geometric correction, and projection — integrated with CI/CD automation and QEMU-based validation.

armasuisse S+T Prime 2021

We also apply our data engineering and ML expertise in non-space domains including logistics optimization, computer vision, and industrial analytics.

Related Services

Planning a Data System?

Whether you're building a new pipeline, deploying ML at the edge, or scaling an existing archive — we'd like to hear about your project.