So... why? If you care about accurate visualization of propogation of light through a complex scene, why not just do a full wave simulation? There are multiple, well studied techniques (Finite Difference Time Domain, Finite Difference Frequency Domain, Finite Element, Method of Moments...). These will give you what the light will 'actually'[1] do, not a guess based off some small sample of training data.
[1] Like with all modeling, there are of course assumptions and approximations. Different techniques have different strengths and limits. Still, these are all well vetted, studied and used by people who care about accuracy.
Prezi used to work this way. (It still may, but I haven't used Prezi in almost 10 years, and it looks like they've gone head first into the AI generated Slide Business)
https://prezi.com/
I used to make entire presentations, systems diagrams, story boards, etc all using scale as a meaningful piece of information. You could go way overboard with it but it was really great. (We used to have a saying "Your Prezi is making me dizzy" for folks that overdid the flying nature)
I'm left handed and had terrible penmanship as a child and all through undergrad. A few things have really helped me.
1. Get pens that write with low friction. As a lefty you're 'pushing' the pen instead of pulling it, and tools with higher friction cause the pen to tilt up in your hand. I found writing with gel pens, a nice fountain pen, or now I use 0.2/0.25 inking pens as my main 'daily driver'.
2. Focus on a few 'problem letters' and get them well sorted. I always struggled with things like 'u' and 'v', '5' and 's', etc. In college I had to make these work, so I spent a lot of time developing a specific style for all of the problem symbols and got *really* good at them.
3. Print vs cursive. I know in school they make you learn cursive, but a lot of the connections between letters decrease readability, and increase complexity for writing them. I no long write in pure cursive, but I have a sort of hybrid print/cursive script. I write many letters in their cursive form as they are faster and flow better, but I never connect letters together. It is a good compromise between speed and readability.
4. Practice, practice, practice. I know this is probably self evident, but it really does take practice. Practice writing letters slowly. It's like martial arts, you go slow and get the movement perfect. Then do it over and over again. That builds the muscle memory better.
[1] Like with all modeling, there are of course assumptions and approximations. Different techniques have different strengths and limits. Still, these are all well vetted, studied and used by people who care about accuracy.