I design and build autonomous AI systems, developer tooling, compiler backends, and multi-agent platforms that push the boundaries of software engineering. High-performance production code across Rust, TypeScript, Python, and C++.
| Project | Category | Stack | Description | |
|---|---|---|---|---|
| ⚡ | MCPlex | Agent Infrastructure | Rust · Tokio · Axum | The MCP Smart Gateway — Semantic tool routing (70–90% token savings), RBAC guardrails, Prometheus metrics, and multi-model support |
| 🔍 | AgentLens | Observability | TypeScript · React 19 · Next.js | Chrome DevTools for AI Agents — Time-travel debugging, real-time per-model token & cost tracking, anomaly detection, MCP-native |
| 🔥 | The Forge | Code Evolution | Python · Vitalis JIT · Multi-LLM | Multi-Agent Code Evolution — 4 LLMs compete head-to-head in an arena, judged by a JIT compiler and bred across generations |
| 🦀 | Vitalis | Compilers & Languages | Rust · Cranelift JIT + AOT | Custom self-evolving compiled language — 345 modules, 6,742 tests, native x86-64, ARM64, and RISC-V execution |
| 🧠 | Gestalt Blueprint | Cognitive Systems | Python · FastAPI · Multi-GPU | Cognitive Architecture & Blueprint Engine — Cross-platform multi-GPU hardware profiling, dynamic VRAM fit, local SLM |
| 🏆 | Neuromantix ARC-AGI-3 | Neuro-Symbolic AI | Python · SMT · Z3 · AST | 88.0% Verified Win Rate (22/25 Games Won) on ARC-AGI-3 with bit-exact deterministic replay |
| 📊 | Neuromantix LiveBench | AI Benchmarking | Python · Symbolic Engines | Official LiveBench submission: 97.0% Reasoning, 100% Data Analysis, 81% Math |
| ✨ | Humaniser | Privacy & NLP | Vanilla JS · Web APIs | Zero-dependency AI writing humaniser with multi-provider Deep Rewrite (OpenAI, Gemini, Claude, DeepSeek) |
| 📈 | Infinity Signal | Quant Systems | Next.js · TypeScript | Professional-grade multi-indicator trading dashboard (15 indicators, 33 instruments, 4 timeframes) |
| 🛡️ | Infinity Security | IAM & Zero-Trust | Rust · Axum · OIDC | Open-source, Rust-native replacements for SaaS — Infinity ID secure IAM (OIDC/OAuth2, MFA, RBAC) + edge gateway |
⚡ MCPlex — The Model Context Protocol Smart Gateway
A high-throughput, asynchronous gateway for the Model Context Protocol (MCP) written in Rust. MCPlex acts as an intelligent intermediary between autonomous AI agents and dozens of MCP tool servers.
|
🎯 Semantic Routing Dynamic vector-based tool selection saving 70–90% in prompt token overhead per agent turn |
🔒 RBAC Guardrails Fine-grained authorization, tool execution policies, and dangerous argument sanitization |
📊 Observability Prometheus metrics, latency histograms, and structured JSON audit logs for enterprise runs |
⚡ Rust-Native Speed Sub-millisecond routing overhead built on Tokio async runtime and Axum web framework |
┌─────────────────────────────────────────────────────────────┐
│ AI Agent (Claude / GPT-4o / Gemini / Local SLM / Cursor) │
└──────────────────────────────┬──────────────────────────────┘
│ MCP Request (JSON-RPC)
▼
┌─────────────────────────────────────────────────────────────┐
│ ⚡ MCPlex Smart Gateway │
│ • Token Optimizer • RBAC Guardrails • Prometheus Monitor │
└──────────────┬───────────────┼───────────────┬──────────────┘
▼ ▼ ▼
[ Local Filesystem ] [ PostgreSQL ] [ GitHub / Slack ]
🔍 AgentLens — Chrome DevTools for AI Agents
An interactive developer cockpit and observability suite designed specifically for agentic LLM architectures. Inspect multi-turn agent execution loops, track per-model token expenditure in real time, and debug tool-calling incidents.
|
Key Features
|
Observability Panels
|
🔥 The Forge — Multi-Agent Code Evolution Arena
An open-source multi-agent code evolution platform where 4 distinct LLMs compete in an arena to generate, optimize, and evolve algorithmic implementations. Solutions are judged by an automated JIT compiler test suite and bred across generations using genetic algorithms.
|
Evolutionary Workflow
|
Engine Highlights
|
🧬 Vitalis — The Self-Evolving Programming Language
A compiled programming language built from scratch in Rust that generates native machine code via Cranelift JIT and AOT. Features an integrated code evolution engine, static effect system, and lifetime analysis. Cross-compiles to x86-64, AArch64 (ARM64), and RISC-V 64.
Source (.sl) → Lexer → Parser → AST → Type Checker → SSA IR → Optimize → Cranelift → Native Binary
↕
C FFI & Python Bridge
|
Compiler Pipeline
|
Runtime Features
|
🧠 Gestalt Blueprint — Cognitive Architecture Engine
Cross-platform multi-GPU hardware profiling, factual VRAM model fitting, and local small-language-model (SLM) synthesis. Dynamically allocates cognitive tasks across available compute tiers (discrete GPUs, Apple Silicon Unified Memory, and CPU fallbacks).
🛠️ Free Developer Tools & Ecosystem (infinitytechstack.uk)
A suite of 10+ production-grade developer tools, interactive academies, and platform extensions — completely free, no sign-up required.
|
Free Developer Tools
|
Interactive Academies & Resources
|
|
Languages |
AI & ML |
Cloud & DevOps |
Security |
Admin |
|
|
|
|
|
|
- ⚡ Expanding MCPlex smart gateway with dynamic semantic routing and vector tool-indexing
- 🔍 Advancing AgentLens for deep inspection and real-time cost tracking of autonomous multi-agent runtimes
- 🧬 Vitalis compiler cross-compilation (x86_64, AArch64, RISC-V) and self-hosted bootstrap pipeline
- 🤖 Scaling multi-agent code evolution and Pareto consensus in The Forge
- 🛡️ Open-source zero-trust security architectures, IAM, and AI agent guardrails
|
🔒 AI Security Audits Threat modelling, prompt injection defence, model supply-chain review from £350 |
🏗️ Architecture Advisory System design for autonomous AI, multi-agent pipelines, Rust toolchains £400 / hr |
⚡ Custom Toolchain Builds Compilers, JIT engines, FFI bridges, SIMD-accelerated libraries Scoped quote |
🛡️ M365 & Zero Trust Entra ID, Intune, Defender, Purview, Sentinel — full stack hardening Scoped quote |
Building systems that build themselves.
100% Open-Source Focus · Rust · Python · TypeScript · C++ · Model Context Protocol (MCP)