Neurosymbolic Systems Engineer

AI/ML Engineer (Neurosymbolic Systems)

We seek an innovative AI/ML Engineer (Neurosymbolic Systems) to join our team to work on pioneering hybrid systems that merge transformer-based deep learning with symbolic reasoning. This role focuses on advancing text/image analysis through architectures that combine neural pattern recognition with rule-based logic, improving accuracy and interpretability in real-world applications.

Role Description

• Contribute to the design of neurosymbolic architectures that integrate transformer/CNN models with symbolic reasoning engines (e.g., knowledge graphs, logic programming, planning engines).
• Implement RAG pipelines using vector/vector-infused databases for context-aware retrieval.
• Optimize transformer models via LoRA and deep fine-tuning for domain-specific tasks while maintaining computational efficiency.
• Develop novel methods to convert neural network outputs into symbolic representations for logical inference and vice versa.
• Evaluate system performance using metrics for accuracy, robustness, and explainability, with adversarial testing to identify edge cases.
• Collaborate on multimodal systems that process text/image/other inputs jointly, leveraging transformer attention mechanisms and symbolic rule engines.

In this position you will build, maintain, and extend bespoke customer-facing systems as well as contribute to our base product platforms. Most of the projects assigned to you are going to require an end-to-end approach; from customer communication, data understanding and data preparation to building functioning and scalable code and AI/ML models and algorithms.

You will problem-solve. Every new system we build is an unexplored territory and a hard nut to crack. Do whatever is necessary to find the best solution for the customer. Do not expect to have high-quality data available, be ready to question everything and build your own way towards the end goal.

Typical Tasks in the Role

• Analyze both structured and unstructured data including images and diagrams creating intermediate structures with utility for comparison, analysis, and question-answering.
• Create customer solutions, from data gathering to display; constructing hybrid AI/ML processing pipelines.
• Deploy, test, validate, and maintain AI / ML models and solutions according to business goals and customer needs.
• Evaluate performance, develop hypotheses for improvement, implement those improvements, and re-evaluate. Rinse-repeat.
• Improve and broaden current ML frameworks and libraries.

Required Qualifications

• Proficiency in PyTorch/TensorFlow and symbolic AI tools.
• Hands-on experience with transformer architectures (e.g., BERT, GPT, LlaMA, ViT, Gemma, etc.) and convolutional networks for multimodal tasks.
• Strong foundation in classic AI/ML: Bayesian networks, graph algorithms, and optimization techniques.
• Experience in MLOps (building and deploying machine learning models in production).
• Experience working with data ingestion and cleaning tools.
• Expertise in model evaluation, including A/B testing, metric design, and bias mitigation strategies.
• Fluency in Python, with experience deploying hybrid AI systems via Docker/Kubernetes.
• Master's or PhD degree in math, computer science, statistics or a related field.

Nice to Have

• Familiarity with neuro-symbolic integration patterns (sequential, parallel, or embedded architectures).
• Experience with automated reasoning frameworks (e.g., IBM’s Project Debater techniques).
• Experience with distributed computing environments and tools.
• Experience with automated system deployment and orchestration for learning (e.g., dstack).
• Publications or projects demonstrating hybrid AI systems for real-world tasks.