AI Initiatives
My Contribution
- I proposed and currently heading up a project for converting the current React MUI client projects into Fimga designs, which in turn can be leveraged by designers as a UI first approach on component development and then converted back into code via Amina.
- After seeing a demo of a POC I did with Cypress Studio, QA brought me into looking at Applitools that is currently going through licensing acquisition.
- Although I've got a license and have used Copilot, it does not quite meet my expectations as such I've found a free alternative with a much more IT trained model called Phind, of which me and the colleagues I've sent have also had great success in the VSCode IDE.
- I can't take credit for the ChatGPT implementation within the Game Distribution Portal that allows for asking the AI bot to query and/or interact with the system data, however I did assist in the debugging and have confidence I can rewrite the same middleware proxy and interface.
- As an personal side project, I've been looking into AI Coding Agents such as Devin, OpenDevin, Codel, SWE-Agent. As these tools evolve, as will the landscape of programming and need to adapt with the industry progress.
- I've proposed creating a chatbot that can query the Confluence wiki through Shakudo, which is currently in costing analysis. In order to ascertain any duplications, get briefs on latest cross-functional work, interpret technical or business language, tutorialise or simplify any docs, get only the specific information you require as snippets and much more.
- Although I've got a license for Midjourney and played around with it, it just doesn't give enough security of IP nor nearly the functionality of any Stable Diffusion GUI and constant copyright claims. The one thing Midjourney has going for it however is ease of use and low barrier of entry. This has lead me down a rabbit-hole of trying all kinds of Stable Diffusion GUI's, such as Stable Diffusion Web UI, ComfyUI, Auto1111 and Fooocus.
- After documenting findings of much experimental work, I've established some custom workflows for non-technical designers, including adapting current IP without training custom models to not infringe upon any copyright agreements.
- I'm currently working on a side project with EfficientSAM that can automatically mask off sections of an image, cut into separate layers for sprite-sheets to plug into Spine or alternative animation tool.