Domain
AI / ML
Applied AI and machine learning — LLM-powered agents, multi-agent systems, RAG pipelines, vector databases, computer vision, and fine-tuned NLP models. Engineering-first: every project is evaluated, measurable, and deployable.
Technology stack
Agent Frameworks
LangChain
Chains & tools
LangGraph
Graph-based agents
AutoGPT
Autonomous agents
CrewAI
Multi-agent crews
Models & APIs
Claude AI
Anthropic LLM
OpenAI
GPT family
HuggingFace
Open models
PyTorch
Training & inference
ML & Training
TensorFlow
Model training
PyTorch
Research models
HuggingFace
Fine-tuning & evaluation
Vector Databases
Weaviate
Semantic search
ChromaDB
Embeddings store
Agents & Multi-Agent Systems
LLM-powered agents with tool use, planning, and inter-agent communication patterns.
Multi-Agent System · LLM Agents
Multi-Agent Task Solver
Core agent loop working, critic integration in progress
A multi-agent system where specialised LLM agents collaborate on complex tasks — planner, executor, critic, and memory agents communicate through a structured message bus.
ReadLLM Systems · Chain-of-Thought
Structured LLM Reasoning Pipeline
Core pipeline working, confidence calibration in experimentation
A modular reasoning pipeline that wraps a base LLM with structured chain-of-thought stages, confidence scoring, and explicit decision boundaries.
ReadRAG & LLM Applications
Retrieval-augmented generation pipelines, AI-powered web apps, and prompt-engineered systems.
RAG · LLM Application
RAG Knowledge Assistant
Pipeline complete, evaluation and chunking tuning ongoing
A Retrieval-Augmented Generation system that ingests documents, builds a vector index, and answers domain-specific questions with cited source passages.
ReadLLM Application · Full-Stack
AI Content Generation App
Core generation flow complete, template versioning in progress
A full-stack AI-powered content generation app with prompt templates, tone controls, output history, and an export pipeline — built on top of the OpenAI API.
ReadMachine Learning & Vision
Trained models, computer vision systems, and NLP fine-tuning — evaluation-first development.
Computer Vision · Embodied AI
Drone Vision & Detection System
Simulation complete, onboard inference optimisation in progress
Real-time object detection and tracking for an autonomous drone platform — YOLO-based inference pipeline optimised for onboard hardware constraints.
ReadNLP · Model Fine-Tuning
Fine-Tuned Intent Classifier
Baseline trained — evaluation and threshold tuning ongoing
A domain-specific intent classification system built on a fine-tuned BERT model, with a full evaluation harness and a FastAPI inference server.
ReadEngineering approach to AI
AI work here is treated as systems engineering — not prompt hacking. Every project has explicit evaluation criteria, measurable quality metrics, and architecture that separates the AI component from the surrounding system. Models are a dependency, not the entire application.