Software that solves real-world problems

Caption Wizard AI

Caption Wizard is a web application built entirely with Next.js that uses DeepSeek language models to generate engaging, on-brand captions for social media and marketing. Users provide a few keywords or context, choose a tone or style, and receive multiple high-quality caption suggestions in seconds.

Case Study

Context and Purpose

Consistent content output is vital for creators and brands, yet writing fresh captions can be time-consuming. Caption Wizard addresses this by automating ideation and wording. DeepSeek powers the caption engine, while Next.js delivers a fast, responsive interface and server-side APIs for reliable generation.

My Role

I led end-to-end development. This included interface design, API architecture, DeepSeek integration, performance optimisation and deployment. I also set up guardrails for safety and quality, so outputs match tone and audience expectations.

Key Features

• Next.js application with server components and API routes for clean separation of UI and model calls.
• DeepSeek integration to generate captions from prompts, keywords, mood and target platform.
• Style controls to switch tone, length and format, with quick re-roll and refine actions.
• Fast responses through request batching, caching hints and graceful retries on provider latency.
• Responsive UI for mobile and desktop with accessible form elements and clear copy actions.
• Quality and safety checks to filter problematic outputs and nudge towards brand-safe language.
• Export and copy workflows so users can paste directly into their social tools.

Impact

• Reduced caption creation time from minutes to seconds for creators and marketing teams.
• Delivered consistent tone and variation across multiple platforms, improving brand coherence.
• Provided a scalable foundation to add multilingual support, templates and scheduling integrations.

Technology Stack

Next.js, React, TypeScript, Node.js APIs, DeepSeek language models, modern CSS (responsive design).

Pluggin Ecosystem

Pluggin Ecosystem is a host and consolidator environment for community provisions across 12 UK regions. It facilitates collaborations between charities, public sector buyers such as the NHS, police and local councils, and businesses that sponsor community initiatives.

Key Contributions

• Planned, designed and developed a complete multi-user onboarding system from scratch using CodeIgniter 4, PHP, JavaScript, HTML, CSS and Bootstrap.
• Architected and implemented role-based user registration, enabling onboarding of three key groups: charities, buyers (public sector organisations) and businesses (sponsors).
• Designed an admin-controlled approval workflow, including notifications, entry validation, secure credential issuance, and enforced password resets upon first login.
• Built a system of unique identifiers for approved users to ensure traceability and governance.
• Developed a bespoke OIDC Identity Provider in PHP, delivering single sign-on (SSO) across multiple platforms.
• Authored a custom WordPress plugin to integrate with the onboarding system, enabling seamless authentication and synchronised user management.
• Integrated WordPress as a community platform, allowing approved users to publish posts, promote causes and showcase business involvement.
• Took ownership of the technical architecture and system design, ensuring scalability, maintainability and security.

Technology Stack

PHP, CodeIgniter 4, JavaScript, HTML, CSS, Bootstrap, WordPress, OIDC (custom implementation).

LEHMR Platform and Related Projects (Medicines Data Map & Kidney Research UK Portal)

As the sole developer, I designed and built the LEHMR platform at the University of Leicester, along with its extensions Medicines Data Map and the Kidney Research UK Portal. These platforms capture the metadata of research projects, ensure compliance with the Health Data Research UK (HDR UK) metadata model, and make projects discoverable to enable wider collaboration.

About the Project

The LEHMR platform supports research discovery by capturing, managing and sharing structured metadata that describes health and medical research projects. Core objectives include:

• Ensuring discoverability of research projects across domains.
• Aligning with the HDR UK metadata standard for compliance and interoperability.
• Providing a secure, user-friendly experience for researchers and clinicians.

Extensions built on this foundation:

• Medicines Data Map – mapping and visualising medicines-related metadata to support research and healthcare decision-making.
• Kidney Research UK Portal – enabling secure discovery of kidney-related research project metadata and supporting collaboration.

My Role

I acted as Technology Lead and sole developer, responsible for end-to-end delivery: planning, system architecture, HDR-compliant metadata model design, backend and frontend development, database schema design, integration and deployment.

Key Contributions

• Designed and implemented the metadata model, ensuring full compliance with the HDR UK metadata model.
• Planned and built the platform architecture using CodeIgniter 4 with clear separation of concerns and extensibility.
• Developed all backend modules for scalable, maintainable performance.
• Created intuitive interfaces with JavaScript, HTML, CSS and Bootstrap for metadata capture and project discovery.
• Designed and optimised the MySQL database for efficient storage and retrieval of structured metadata.
• Implemented secure authentication and role-based access controls suitable for sensitive research environments.
• Extended the LEHMR core to power Medicines Data Map and the Kidney Research UK Portal without code duplication.
• Ensured strong data governance and compliance across all implementations.

Impact

• Delivered a metadata-driven discovery platform underpinning multiple health research initiatives.
• Achieved HDR UK compliance, improving interoperability with wider research infrastructures.
• Provided a scalable and reusable solution that maximises return on development.
• Enabled researchers and clinicians to find, explore and connect with relevant projects, driving collaboration.

Technology Stack

PHP, CodeIgniter 4, JavaScript, HTML, CSS, Bootstrap, MySQL.

DUC / DUCEPND / DUCEJPRD — Digital Use Conditions for Research Data Governance

I served as the sole technology lead, architect and developer for the DUC platform and its extensions DUCEPND and DUCEJPRD. DUC turns consent and data use conditions into a machine-readable, searchable model that organisations can apply to datasets and biosamples. DUCEPND adapts DUC for the neurodegeneration community, and DUCEJPRD focuses on rare disease contexts using the Common Conditions of Use Elements (CCEs). I did not design the original use-condition framework, but I translated it into a practical, robust software system and shaped a usable schema for real-world workflows.

Context and Purpose

The Digital Use Conditions (DUC) approach provides a structured way to express permissions, obligations and restrictions around research data. Profiles describe resources and the conditions under which they may be used. Background and overview: About DUC (PDF).

Extensions built on this foundation:

• DUCEPND — adapted for the European Platform for Neurodegenerative Diseases (EPND) to capture and expose cohort, dataset and biosample metadata relevant to neurodegeneration research.
• DUCEJPRD — tailored for the European Joint Programme on Rare Diseases, incorporating the CCE vocabulary to make complex use conditions clear and consistent. CCE context: CCE explanations (PDF).

My Role

As the sole developer, I owned planning, architecture, schema design, backend and frontend development, database design, security, deployment and ongoing refinement. I worked closely with domain experts to ensure the platform reflected real data-governance needs while remaining straightforward to use.

Key Features and Contributions

• Schema and Metadata Model — implemented the DUC conceptual model in a practical database-backed schema, including condition term, rule, scope and parameter structures.
• CCE Integration — incorporated Common Conditions of Use Elements to break down complex policies into usable, non-directional building blocks for DUCEJPRD.
• Profile Creation Flows — built guided interfaces to help curators compose, validate and preview DUC profiles, with export options for downstream use.
• Search and Discovery — delivered filtering and browsing tools so researchers can quickly understand conditions before requesting access.
• Security and Governance — added role-based access, auditability and validation to support responsible data stewardship.
• Domain Extensions — extended the DUC core for DUCEPND and DUCEJPRD with fields and UI patterns specific to neurodegeneration and rare disease consortia.
• Publication Support — contributed the technical implementation referenced in the associated peer-reviewed research, aligning software behaviour with published methodology.

Impact

• Made abstract consent and use conditions actionable through a clear, machine-readable platform.
• Enabled EPND and rare disease stakeholders to catalogue resources with transparent, consistent conditions.
• Improved discoverability and trust by showing researchers exactly how data may be used.
• Aligned with FAIR principles, supporting interoperability across projects and infrastructures.

Technology Stack

PHP, CodeIgniter 4, JavaScript, HTML, CSS, Bootstrap, MySQL.