Building the future of energy. Together.
CAELUS is a deep tech startup with a clear mission: to bring artificial intelligence to the heart of nuclear infrastructure.
Our culture is based on innovation, collaboration and impact.
We are a multidisciplinary team: engineers, designers, researchers. We grow with honest feedback, ownership and attention to detail.
Why work at CAELUS
An environment that values people, ideas and growth.
Remote-Friendly & Hybrid Work
Real flexibility, shared responsibility
Clear Growth Paths
Evolving roles and internal mentoring
Continuous Training
Annual budget for courses, conferences, certifications
Sustainable Impact
Work on technologies that change the energy sector
CAELUS Research Program 2025
CAELUS runs one of Europe's most advanced research programs at the intersection of artificial intelligence, simulation, regulatory automation and deeptech product engineering. We collaborate with Master's students and early-stage researchers to develop thesis projects that directly contribute to our real product roadmap.
All projects are designed to be technically challenging, highly impactful and co-supervised by CAELUS engineers together with university advisors. Below you can find a selection of current thesis opportunities.
All thesis projects are conducted in English and can be structured as Master Thesis, Research Internship or combined format, depending on university requirements.
Track 1 — AI & Simulation
AI Reasoning Engine for Regulatory Automation
Design and develop an AI reasoning engine that assists in interpreting and structuring complex regulatory documents for the nuclear and energy sector. The project focuses on building an LLM-based pipeline that can extract requirements, identify dependencies, and suggest structured representations suitable for automation. The student will work with modern transformer architectures, prompt-engineering strategies and evaluation datasets defined by CAELUS. The outcome is a research prototype that demonstrates how AI can support more reliable and transparent regulatory workflows.
3D Digital Twin Generation via NeRF & Scene Reconstruction
Explore Neural Radiance Fields (NeRF) and related methods to reconstruct realistic 3D digital twins of industrial environments from video and image data. The goal is to build a pipeline that ingests camera footage and outputs interactive 3D scenes that can be used in CAELUS SiteScope and XR experiences. The student will benchmark different NeRF variants, study trade-offs between quality and performance, and integrate the results into a simple viewer. This is a highly visual and technically advanced project, ideal for students interested in computer vision and graphics.
Physics-Aware Denoising for Technical Sensor and Simulation Data
Develop a physics-aware denoising model that cleans technical data (e.g. time series, simulation outputs, thermal maps) without destroying the underlying physical structure. The goal is to explore neural architectures that incorporate physical constraints, so that the output remains realistic and usable for downstream analysis. The student will generate synthetic noisy datasets, compare classical filters vs deep learning approaches, and evaluate the impact on forecasting and anomaly detection. The final result is a denoising component that can be integrated into CAELUS Engine and Maestro.
Model Drift & Real-Time Explainability for AI in Industrial Settings
Design a monitoring framework to detect when AI models used in industrial or regulatory scenarios start to drift from expected behavior. The student will implement metrics, dashboards and alerting mechanisms that track data distributions, prediction stability and confidence changes over time. The project also includes experimenting with explainability techniques (e.g. feature importance, example-based explanations) to help users understand why a model is changing. The outcome is a research prototype that can plug into CAELUS MLOps pipelines to keep models trustworthy over time.
Track 2 — Platform Engineering & Data
CAELUS OS — Modular Runtime Architecture for Enterprise Deeptech Platforms
Design a conceptual "CAELUS OS" runtime architecture that coordinates multiple AI services, data pipelines and user-facing applications under a single, modular core. The thesis focuses on defining boundaries between modules, communication patterns, configuration handling and lifecycle management for services. The student will analyze best practices from cloud-native and microservice ecosystems, and propose a reference architecture for a deeptech platform like CAELUS. The outcome is a system design and a proof-of-concept implementation that demonstrates how modules can plug into a shared runtime.
Multi-Tenant Architecture & Role-Based Access Control for CAELUS Platform
Define and prototype a multi-tenant architecture that allows CAELUS to serve multiple organizations (utilities, engineering firms, regulators) in a secure and isolated way. The student will study different tenancy models, design a role-based access control (RBAC) scheme, and propose a permission structure that can support complex organizational hierarchies. The thesis includes building a small backend prototype with tenants, spaces, roles and permissions, and demonstrating how this architecture can scale for enterprise usage.
Data Governance & Metadata Lineage Framework for Enterprise AI Systems
Design a data governance and metadata lineage framework for CAELUS, focusing on how datasets are created, transformed, validated and consumed across the platform. The student will define entity models for datasets, pipelines, transformations and models, and propose how lineage information should be stored and queried. The goal is to provide a clear "map" of how data flows through the system, supporting transparency, compliance and debugging. The output is both a conceptual model and a small prototype showing lineage tracking in action.
Automated Testing Framework for Multi-Module Deeptech Platforms
Create a testing framework that can run automated checks across multiple modules of the CAELUS platform (backend services, APIs, AI components and important UI flows). The thesis involves defining testing layers (unit, integration, end-to-end), designing reusable test fixtures, and integrating tests into a CI pipeline. The student will also explore how to validate non-deterministic AI behavior using tolerance ranges and reference outputs. The outcome is a practical test harness that can be extended by future CAELUS developers.
Track 3 — Market Intelligence & Strategy
Competitive Intelligence Radar for Global Deeptech & Industrial AI
Design a competitive intelligence "radar" that monitors activity in deeptech, industrial AI and digital twin companies worldwide. The student will collect and structure public data (websites, research papers, news, funding announcements) and build a simple system that classifies and tracks relevant players over time. The thesis explores clustering, tagging and basic analytics to help CAELUS understand where the market is going and which niches are still underserved. The goal is to deliver a living dashboard that can support strategic decisions.
Pricing Strategy & Econometric Modeling for B2B Deeptech Platforms
Develop a pricing strategy for a modular B2B deeptech platform, using basic econometric and business modeling techniques. The student will analyze different revenue models (subscription, usage-based, per-seat, per-project), study how value is perceived by different customer segments, and propose a pricing framework for CAELUS products. The thesis includes building simple financial scenarios and sensitivity analyses that show how pricing choices affect growth and sustainability.
Customer Adoption Metrics & Health Index for Enterprise Clients
Define a quantitative "health index" that summarizes how well a customer is adopting and benefiting from CAELUS products. The student will identify meaningful usage signals (feature adoption, login frequency, project activity, support tickets), design engagement scores and thresholds, and propose dashboard views for Sales and Customer Success teams. The goal is to detect early which customers are thriving, which are at risk, and where proactive support could make a difference.
Strategic Valuation Model for Multi-Product AI Platforms
Build a strategic valuation model tailored to multi-product AI platforms operating in complex B2B markets. The student will combine elements of traditional startup valuation (TAM/SAM/SOM, revenue multiples, scenarios) with platform-specific factors like ecosystem effects, data moats and product modularity. The thesis focuses on defining the right variables and presenting them in a clean, decision-friendly way for investors and internal stakeholders. The output is a structured valuation framework that can be reused as CAELUS grows.
