Introductions
Read my Blog Connect on LinkedIn Follow on X (Twitter)
Who I Am¶
An experienced technical consultant specialising in AI Solutions and Generative AI implementation, with an academic AI background. A Professional Member of the Institute of Analysts and Programmers holding an dual honours degree in Artificial Intelligence and Computer Science from the Russell Group University of Birmingham.
I help teams, companies, and organisations:
- Build AI prototypes and solutions
- Solve problems using AI
- Technical reviews and evaluation
- Implement AI in a future-proof manner with upgrades in mind
- Advise on state-of-the-art AI tools and knowledge
I have carried out technical reviews of AI Book Publications for Springer Apress. Also, technical reviews of Edge AI University course material and lessons as part of a Global Education Project to help educate students in EdgeAI with Imagination Technologies, Madrid University and Peking University.
Services¶
If you are challenged with AI, I can help you:
- Understanding AI capabilities and limitations: What is possible now and likely in the near future.
- Retrieval-Augmented Generation (RAG): Implementing RAG with hybrid vector and semantic search with re-ranking. Including text, tabular data and images. Evaluating Embedding models.
- Prompt Engineering: Techniques to craft effective prompts for Generative AI.
- Generative AI hallucination: Techniques to reduce hallucination and optimise consistance.
- Quality Control and Evals (Evaluations): Maintaining consistency and reliability in Generative AI. Understanding how to evaluate and have confidence in the content your Generative AI produces, including after changes. Implementation of LLM as a Judge evaluations.
- Using OpenAI Services and APIs: Experienced with OpenAI's APIs and Azure OpenAI Service, Azure AI Search (formerly Cognitive Search).
- GPT models, Anthropic, Llama LLMs: Experienced with Large Language Models (LLMs), including Open-weight/ Open-source models.
- Fine-tuning models: Fine-tuning both LLMs and Image models, LoRa training. When to fine-tune and when not to.
- Generating high quality synthetic content: Using LLMs and hybrid RAG to generate accurate content.
- Agents: Building prototytes of agents, multi-agent systems in LangGraph.
- Deep learning based Computer Vision: Computer Vision Model arcitecture and training.
- Image Generation: Understanding how prompts, diffusion models and latents are used in Generative AI, including Stable Diffusion.
Interesting Facts¶
- Used Speech synthesis first in Amiga basic.
- First programmed with NLP using Prolog in 1998.
- Trained my first neural network to make predicitions on tabular census data in 1999 at university.
- Used Genetic Algorithms to evolve Cellular Automata for Computer Vision for my thesis in 2000, which has been cited.
- Took part in Jeremy Howard’s challenge in the last fast.ai part 2 course and topped the leaderboard in every epoch category, getting the highest classification accuracy on the Fashion-MNIST dataset in only 5, 20 and 50 epochs of training - using a non-pretrained model.
- One of the first cohort of 1000 SolveIt students.
Skills and technologies¶
- Prompt engineering for Language Models
- LLMs, SLMs, VLMs, RLMs
- RAG, advanced
- AI Tools: LangChain and LlamaIndex
- AI Agent frameworks: LangGraph, Smolagents, LlamaIndex Workflows
- PyTorch, Python, .Net, C#
- Quantisation of Edge AI models and testing models running on FPGA implementations of Neural Network Accelerators
- Synthetic data generation
LLM an VLM Model Families¶
Worked extensively with the Gemini model family and OpenAI model family.
Image Generation Techniques & Technologies¶
- Prompt engineering for Image and Video generation
- GenAI Image generation
- Diffusion models
- Stable Diffusion
- Style transfer
- LoRa training
- Super resolution
- Colourisation