Projects
The Architect
Production (PyPI)Public • Open-Source
- Autonomous AI development lifecycle framework empowering agentic coding tools to plan, execute, retry, review, and retain knowledge without human intervention.
- Utilises persistent project memory stored in ARCHITECT.md to accumulate decisions, constraints, and lessons across sessions, ensuring continuity.
- Implements multi-signal completion detection and circuit breakers to avoid agent stalls and hallucinated outputs.
- Supports headless and continuous integration (CI) modes alongside live tmux dashboard for flexible deployment and monitoring.
- Available on PyPI with seamless compatibility to Claude Code and OpenCode environments.
MHIL
Concept & Early Prototyping (TRL-5)Semi-Stealth • Non‑Profit
- Machine Human Intelligence Lab (MHIL) is a non-profit initiative creating comprehensive benchmarks and evaluation protocols that transcend narrow LLM performance metrics.
- It develops human-aligned test batteries targeting agents and composite AI systems with capabilities spanning memory, planning, robustness, and autonomous problem-solving under uncertainty.
- MHIL offers open evaluation matrices for 'pre-superintelligent' systems, enabling standardized, architecture-agnostic comparison across models and vendors.
- The framework sets future readiness standards with reproducible scorecards, capability tiers, and safety-critical thresholds tailored for responsible AI deployment.
MazeByte
SOTA Proof of Concept (PoC)Semi-Stealth
- Proof-of-concept demonstrating a fully autonomous AI-driven ETL agent capable of self-maintaining data pipelines without human oversight.
- Incorporates infant-like reinforcement mechanisms with reward and pain signals coupled with resource-seeking behaviours for adaptive learning and system resilience.
- Automatically generates pseudo-code modules encompassing memory management, search algorithms, optimisation routines, and orchestration to perform end-to-end ETL automation.
- Built on the proprietary MBI model, MazeByte aims to eliminate manual data engineering workflows by treating data systems as adaptive, agentic entities.
SightBit
SaaS (TRL-9)Public
- Production-grade computer vision SaaS platform delivering real-time aquatic safety monitoring through drowning prediction, human and vessel detection, and flood risk alerting.
- Operates exclusively on standard CCTV feeds without requiring specialised sensors, edge devices, or site-specific calibration.
- Employs a proprietary panoptic segmentation model enhanced by an advanced augmentation pipeline, significantly improving small-object recall and detection robustness.
AdaptiveBridge
Production-ready (TRL-8)Public • Open-Source
- Open-source Python library designed to reinforce machine learning model robustness against missing features during inference in production environments.
- Utilises configurable thresholds for feature importance, correlation, and accuracy, supporting customised accuracy logic and automated distribution adjustments.
- Provides a drop-in estimator interface requiring a single fit operation followed by straightforward transformation calls, enabling easy integration into existing ML pipelines.
- Manages complex feature dependencies including mandatory, deviation, and leveled relationships to optimise model performance under incomplete inputs.
- Delivered with comprehensive documentation, unit tests, continuous integration, and a permissive open-source license.
Jhive
Production (TRL-7)Public • AIforGood
- Django-based production web application leveraging the OpenAI API to detect antisemitic content through real-time natural language classification.
- Features a detailed reporting interface to support operational content moderation.
- Exposes RESTful API endpoints with rate-limiting controls for seamless integration.
- Employs rigorous input validation and sanitisation protocols to safeguard against injection and security vulnerabilities, ensuring reliable and responsible deployment.