The Lab
Our research and development hub where we experiment with cutting-edge technologies, contribute to open-source, and share our learnings with the developer community.
Our Open-Sourcing Strategy
Building in Public
We believe that the best technology emerges from collaborative development and transparent iteration. Our open-source approach isn't just about giving back—it's about building better products through community engagement and shared innovation.
Core Principles
- Research-First Development: Every project starts with thorough research and experimentation, making our findings available to the community.
- Production-Quality Standards: Our open-source tools meet the same quality standards as our commercial products, with comprehensive documentation and testing.
- Community-Driven Evolution: We actively engage with users, incorporating feedback and contributions to improve functionality and usability.
Current Focus Areas
We're evolving into a developer-first AI studio, building tools and platforms that empower engineers to work smarter with large language models.
Developer Platforms
Platforms and APIs that help developers transform complex challenges into simple solutions.
Flagship product: Contextus, your AI scalpel that cuts token bills by 40% without upgrading your model. We help your AI prioritize what matters & discard the rest. Contextus plugs in whether you're vibe coding prototypes or building with agent pipelines, making your LLM calls run smaller, faster, and truer to your intents.
AI/ML Tooling
Specialized models and tools for exception handling, reconciliation, and next-gen AI experiences.
Includes open-sourced ML models (DistilBERT-Reconciler, Fail-Forecaster, Stress-Flagger) available on Hugging Face.
Commitment to publishing research, white papers, and developer resources to push AI forward.
Contributing Guidelines
We welcome contributions from developers at all levels. Whether you're fixing bugs, adding features, or improving documentation, your input helps make these tools better for everyone.
Getting Started
Each repository includes comprehensive setup instructions, contribution guidelines, and development documentation.
Open Source Models
Research models and tools available on Hugging Face
Fine-tuned DistilBERT on 3.2M labeled post-trade break descriptions + resolution actions.
Predicts next-day settlement-fail notional for US Treasuries & corporates using XGBoost + LightGBM.
Flags days where settlement fails are in the top-10% of historic values for proactive remediation.
White Paper
Musodza, K. (2025). Bond Settlement Automated Exception Handling and Reconciliation. Zenodo.
Projects
From production-ready platforms to experimental lab projects, explore our tools and contributions to the developer community.
Loading projects...