Hello! I'm Vincent Haywood, an implementation-focused AI and automation consultant. I help companies and agencies move beyond experimentation and actually deploy AI — through intelligent workflows, automation, and micro-SaaS products built for real operational use.
Projects
A collection of client work, click to expand
Podcast → Central Brain
Turns long-form audio or video into a searchable knowledge base and ready-to-publish content automatically.
An end-to-end automation that ingests long-form audio or video and turns it into a searchable knowledge base and platform-ready content. The system transcribes, structures, and embeds conversations into a RAG database, then generates summaries, social posts, descriptions, and clips automatically. This removes manual editing and drastically reduces content turnaround time from days to minutes.
Centralised Agency Knowledge Brain
A RAG-powered internal knowledge system that lets teams query their own data in natural language.
A centralised knowledge system that unifies documents, conversations, reports, and institutional knowledge into a single, queryable source. Teams can ask natural language questions and receive context-aware answers grounded in their own data, reducing time spent searching and preserving organisational memory.
AI Enablement & Adoption Programme
Enabled teams to adopt AI safely, confidently, and effectively.
Designed and delivered an enablement programme combining training, internal documentation, practical use cases, and guardrails. The focus was on building confidence and competence across roles, ensuring AI became a day-to-day capability rather than a novelty.
Experiments
Click to expand
AI Influencer & LoRA Training Pipeline
Custom LoRA training workflows for creating consistent, realistic AI-generated characters.
An experimental pipeline covering dataset generation, LoRA training, and controlled output across styles and formats. Focused on realism, repeatability, and understanding the practical limits of synthetic media systems.
RAG Quality & Failure Mode Experiments
Testing chunking strategies, retrieval depth, prompt structure, and embedding approaches to improve answer quality.
An applied experiment focused on improving the reliability of Retrieval-Augmented Generation systems by testing how ingestion, chunking, embeddings, retrieval depth, and prompt structure affect answer quality. The work explores common failure modes such as hallucination, weak grounding, and irrelevant context, and uses those findings to design guardrails and patterns that produce more accurate, trustworthy, and scalable RAG systems in real-world workflows.
Client onboarding
Remove the long process of info gathering and folder setups for new clients
Allow new clients to fill in a simple form, this then powers new client creation on Xero, folder creation on Google drive, project plan and proposals in Notion. One form, multiple jobs done.