Project: easyAI-MRO

This is a log of bringing the product to market.
01-08-2025: Decided to start a new project that build and brings to market a product using AI for the MRO market (Civil Aerospace and Defence) The ideea will be to build a product for the MRO market using AI to aid inspectors in cataloging and capturing images and data of the defects found at the stage of visual inspection. then in the back-end the images are processed using another AI for defect detection (dents, scratches, inclusions, burn, corrosion and so on... ) and it automatically catalogues and builds the report suggesting next steps.

Business Case
At this point in time I am thinking of making it open source for whoever would like to play / install it they can do it , however as it is higly specialized they could pay for customized service and implementation. No point in keeping the code secret as AI evolves so fast that the only way to keep ahead is just to keep innovate. Payment will be subscription based and intial installation involves having onw local hardware/server

Flow/Feutures
Keeping it easy at this point in time- 2 steps process:

• 1. Capture - primarely frontend capture the images/ videos with audio and tags (build in tagging mechanism) with minimal feedback to the inspector/user - Using an AI like OpenAI for automatic closed captcha and for automatic chaptering and capture of the action. Possible use of other sensors like accelerometers and camerasto capture more data.
• 2. Data processing and reporting - using AI to catalogue and do image segmentation/ defect detection in the back end and to generate reports.
• other possible feutures might be image /video stiching for long or round parts.
• connection with other databases like SAP or ERPs

02-08-2025:
Have been looking into the different technologies could use to build the MVP app myself up to 80% of proof
Technologies: Tech stack
I can build this in Python Django, keep it simple for the MVP ( no API between front and back end at this point in time). 1. Capture - been testing: can use JS for now and then later on swithc to Java or more advance. For now use JS and PWA to capture image/video / audio and tags and send it to back end for step 1: data capture and chaptering. Here is have several options for an AI to do chaptering, transcribing and tagging

Tech Stack (MVP only - as of today):
• - Python Django/HTMX: core
• - Front end:JS and PWA for capturing content : TEST completed succesfully
• - AI transcript and chaptering:
• - TESTING... Whisper by OpenAI
• - AI defect detection:
• - TESTING... getting my hands dirty, training own model and implement YOLOv8 or YOLOv11 - unlikely - lack of dataset
• - TESTING... Segment Anything or Sam2 - segment anything 2 - by Meta - github - credit and thx for the suggestion: Albert Phang