real problems, shipped
Inbal Waller
Learning & AI Enablement · Hands-on builder
These are real projects from my own business and from a nonprofit training program I coordinated. In each one I defined the problem, directed Claude and other AI tools to build the solution, checked the output, and put it in front of real users. None of this is a mock-up.
STEP 1Diagnose
Find what is actually blocking people before building anything.
STEP 2Build with AI
Claude does the heavy lifting. I set the direction and the constraints.
STEP 3Verify
Check every output against the source data before it reaches anyone.
STEP 4Ship and iterate
Real users, real feedback, then the next version.
Featured case study · Learning design
Doodling, Just Begin
A self-paced video course for adults with no drawing background who want a calm creative break. The main barrier is not skill. It is the fear of the blank page and the belief “I’m not talented.”
- Learning goal
- By the end, a learner completes a full colored composition on her own, using page division, four repeat patterns and four botanical elements.
- Design decisions
- Start with the barrier. The first module removes the fear of the blank page by working on small A5 and A6 formats and dividing the page before drawing anything.
- Scaffolding. Four page divisions, then four patterns, then free practice. Four botanical elements, then a full composition. Finally color, which combines everything into one finished piece.
- Microlearning. 25 short lessons of 3 to 10 minutes each, built for watching on a phone.
- A low entry bar. All you need is a pen and blank paper.
- A reflective close: a final lesson on taking the calm forward.
- Production
- I filmed and edited every lesson in Premiere and recorded the narration. The lessons are hosted on Vimeo, the course is built in Tutor LMS, and it is connected to a one-page checkout.
- Outcome and iteration
- Launched in 2026. Together with a follow-on course, it has 25+ students. After launch I removed friction from the checkout, added a login FAQ, released a follow-on course and created a bundle.
Free sample: a beginner lesson in the same teaching style (in Hebrew).
What learners say
“I thought I would never have the patience. But one small thing you said at the start, that a big empty canvas is intimidating so it’s better to begin on very small pages, changed it for me. Now I wait every evening for my time with the pages and pens.”Participant, translated from Hebrew
“It brought calm to the soul, together with practical tools.”Participant, translated from Hebrew
Group 1
Learning and enablement
KhalasTech program · Halasartan nonprofit · 2026Mentor matching and program operations
- Problem
- Fifteen mentors had to be matched with participants and briefed. The data was spread across a form tool, a participants file, a pairings file and a contacts list.
- What I built
- Cross-checked all four sources and caught errors before anything went out: email addresses reversed by right-to-left text extraction, an ambiguous row, and a participant no longer in the program. Then a matching sheet with color-coded need categories, 15 personalized mentor emails with a print-ready pack, and a three-sheet effectiveness survey with response tracking and an auto-calculating summary.
- Result
- Every mentor was briefed from verified data, and the program has a built-in way to measure itself.
- Claude
- Excel
- Word
Inbalika digital courses · 2026A production pipeline for video courses
- Problem
- I produce video courses alone. Drawing, filming and editing take most of the energy, which leaves little time for structure, scripts and platform setup.
- What I built
- A clear split of work. I keep the drawing, filming, editing and voice. AI drafts the lesson structure, writes narration scripts from my timestamped notes, writes lesson texts and downloads, and builds the course in the LMS. For the newest course, AI broke a reference video into construction stages, which became three shorter lessons.
- Result
- Two courses launched in 2026 with 25+ students, and a repeatable workflow for the next one.
- Claude
- Premiere Pro
- Audition
- Vimeo
- Tutor LMS
Inbalika digital courses · 2026Self-serve learner onboarding
- Problem
- Learners bought a course and then could not log in. The checkout also had a confusing separate registration step.
- What I built
- A one-page checkout that handles payment and enrollment in one step. New login instructions in the enrollment email. A bilingual FAQ that answers pre-purchase doubts and post-purchase login questions with screenshots, marked up for search engines, and linked from course pages, checkout, the site footer and the enrollment email.
- Result
- The login answer now meets learners at every point where they used to get stuck.
- Claude
- WordPress
- CartFlows
- Tutor LMS
- Yoast
Inbalika social and Etsy content · 2026Automated short-video editing
- Problem
- Turning raw footage of my drawing into short social videos took a long manual edit for every clip.
- What I built
- A scripted edit: vertical 9:16 crop, normal speed at the start and the final reveal, a sped-up middle, and handwritten-style captions that appear as if being written. A second version pairs drawing clips with AI product videos and English captions for the Etsy shop.
- Result
- Adopted as the standing house style. I add music and publish.
- Claude
- ffmpeg
- Python
- Kling AI
Group 2
Business operations and systems
Inbalika paid campaigns · 2026Fixing ad purchase tracking
- Problem
- Course sales were happening, but the ad campaign was not seeing them. The platform was optimizing on the wrong signal, and ad spend was higher than course revenue.
- What I built
- Traced the site’s tracking and found two different pixels firing, with the campaign tied to the one that received no purchases. Documented the full setup for rollback first. Then moved the ad account to the right business portfolio, connected the working pixel, rebuilt the campaign and three custom audiences on it, and launched a retargeting campaign that excludes past buyers.
- Result
- The campaigns now learn from real purchase data.
- Claude
- Browser automation
- Meta Business Suite
inbalika.com · 2026Legal and accessibility review
- Problem
- I was not sure the site’s forms and policies met anti-spam and accessibility requirements.
- What I built
- An audit that found 11 forms mixing required consent with marketing consent, and a hidden third-party email recipient. Split consent into a required box and an optional, unchecked marketing box. Updated the privacy policy and published an accessibility statement. Fixed button contrast to a 4.5:1 ratio, labelled icon-only buttons, made the contact widget keyboard-operable and added missing page headings.
- Result
- Every form now matches the site’s own privacy policy, and the brand buttons pass WCAG AA contrast.
- Claude
- WordPress
- Elementor
Inbalika workshops · 2026Landing pages by audience
- Problem
- One workshop page tried to speak to organizations and to private hosts at the same time.
- What I built
- Two separate landing pages, one for organizations and one for private groups, plus a hub page and menu that route each visitor to the right one. Replaced AI illustrations with real workshop photos and anonymized 15 testimonial screenshots by blurring names, faces and phone numbers. Set up search settings for each page.
- Result
- Each audience lands on a page written for it.
- Claude
- WordPress REST API
- Elementor
- Yoast
Inbalika brand · 2026A brand design system
- Problem
- Every new design started with re-explaining fonts, colors and style.
- What I built
- A single reference with color and type tokens, components, logo assets and a writing-voice guide. It was built from the live website and my published designs, and it flagged the site buttons that failed contrast.
- Result
- One source that both I and AI tools use for all branded work.
- Claude
- Canva
Inbalika planning · 2026From work plan to calendar
- Problem
- A one-month work plan sat in a document and did not turn into daily action.
- What I built
- An automated pass that turned the plan into 39 calendar blocks over four weeks, following a fixed weekly rhythm. It worked around every existing appointment without overwriting any of them, and flagged the one conflict for me to decide.
- Result
- The plan now lives in the calendar.
- Claude
- Google Calendar
This page grows as new projects ship.Last updated 15 October 2026