Applied AI Researcher— Sovereign AI Research

At Dream, we redefine cyber defense vision by combining AI and human expertise to create products that protect nations and critical infrastructure. This is more than a job; it’s a Dream job. Dream is where we tackle real-world challenges, redefine AI and security, and make the digital world safer. Let’s build something extraordinary together.

Dream's AI cybersecurity platform applies a new, out-of-the-ordinary, multi-layered approach, covering endless and evolving security challenges across the entire infrastructure of the most critical and sensitive networks. Built as part of a broader sovereign AI platform, our technology is designed to operate in on-premise, private cloud, and air-gapped environments, enabling nations to maintain full control over their data, infrastructure, and AI capabilities. Central to Dream's proprietary Cyber Language Models are innovative technologies that provide contextual intelligence for the future of cybersecurity.

At Dream, our talented team, driven by passion, expertise, and innovative minds, inspires us daily. We are not just dreamers, we are dream-makers.

The Dream Job:

Nations are waking up to a hard truth: critical intelligence infrastructure built on hyperscaler black boxes isn't a solution it's a dependency. The DREAM Sovereign AI Research Group exists to answer that differently.

We're not fine-tuning what already exists. We're rethinking the models architecture from the ground up modular, composable, and built with compute governance as a first-class design constraint, not an afterthought.

We operate under real-world constraints. The interesting problems live at the intersections of disciplines. That's where we operate.

This is a hands-on research role embedded within a team of senior researchers. Day to day includes training models, running benchmarks, synthesize data. The researchers you'll work alongside will challenge you technically and invest in your growth.

The Dream-Maker Responsibilities:

Open Research Tracks

Familiarity with at least one is expected:

• Computer Vision: object detection, segmentation, multimodal grounding, vision-language models, contrastive and self-supervised representation learning, low-resource and few-shot visual recognition.
• NLP / Speech: LLMs, NERs, relation extraction, span-based and generative IE, semantic textual similarity, multilingual and cross-lingual transfer.
• Reinforcement Learning: MDPs, POMDPs, model-based and model-free RL, Online Offline methods, reward modeling, sim-to-real transfer, compute-aware planning.
• Graph Learning: GNNs, graph clustering, community structure, generative methods, knowledge graph embeddings, dense and sparse semantic retrieval.
• Optimization: convex and nonconvex optimization, constrained and Lagrangian methods, combinatorial and integer programming, knowledge distillation (response, feature, and relation-based), test-time optimization, Bayesian optimization, resource-aware inference.
• Representation Learning: contrastive learning, self-supervised and unsupervised pre-training, disentangled representations, metric learning and embedding spaces, cross-modal and multimodal alignment, meta learning (hypernetworks), transfer learning and domain adaptation, probing and interpretability of learned representations, world models.
• Neurosymbolic AI: neuro-symbolic integration, differentiable theorem proving, inductive logic programming (ILP), probabilistic soft logic (PSL), causal inference and structural causal models (SCMs), programmatic and compositional reasoning
  
  

• Train and evaluate models across research tracks, iterating fast while documenting rigorously.
• Build and maintain benchmarking pipelines and evaluation suites.
• Curate, structure, and preprocess datasets; contribute to synthetic data generation workflows.
• Run ablations and controlled experiments to support research hypotheses.
• Reproduce and stress-test results from recent literature relevant to the group's work.
• Collaborate across tracks and with engineering teams through to production handoff.
  
  

• MSc in Computer Science, Electrical Engineering, Mathematics.
• Strong academic record with hands-on experience / thesis research.
• Proficiency in Python and at least one deep learning framework (PyTorch preferred).
• Comfort with the full data lifecycle: sourcing, structuring, cleaning, and transforming raw data into training-ready assets