I use AI to turn small operational problems into working tools.

Tool 01 · Work project

Referral tracking dashboard

Context & problem

Objective: show how many visitors a regulation content site sends to a key service page, next to the site's Google search traffic.

Challenge: content was judged on traffic alone, with no simple view of the business it drives.

Approach

• Defined what to measure: clicks to the service page, clicks to the rest of the product site, and search clicks for comparison.
• Chose a lightweight setup: live data without the dashboard calling external APIs.
• Made it self-serve: built-in connection checks, step-by-step setup guidance and a README for the developer who deploys it.

1. Link clicks on the content site
2. Tracking script
3. Database
4. Dashboard

Search Console data is synced into the same database once a day.

Results

• Delivered: a working live dashboard and setup guide, ready for the team to deploy.

Tool 02 · Personal project

Search Terms Cleaner

Context & problem

Objective: make the Google Ads search-term review faster and safer.

Challenge: finding wasted spend by hand is slow, and changing an ad account by mistake is costly.

Approach

• Clear rules flag terms with zero conversions, high cost, brand or competitor names, research intent, locations and near-duplicates.
• Suggestions for negative keywords, with the right match type and a caution when a term contains the brand.
• Safe by design: read-only, with adjustable rules and a CSV export for review.

1. Search terms
2. Rules
3. Flags
4. Suggested negatives
5. CSV for review

Results

• Delivered: a working, open-source review tool with demo data, available on GitHub.

Tool 03 · Personal project

TJS: a speech therapy clinic platform

Context & problem

Objective: one workspace for a speech diagnostics department, a speech therapy department and university clinics where supervisors allot cases to students.

Challenge: the work runs across several roles and handoffs, and it handles sensitive health data, so privacy had to be designed in from the start for GDPR and India's DPDP Act.

Approach

• Mapped the workflow from my clinical background: register and assess, refer to therapy, allot to a therapist or student, then supervisor review of lesson plans and reports.
• Need-to-know access by role: supervisors, diagnosticians, therapists and students each see only what their role needs.
• Consent per purpose: nothing pre-ticked, in five languages with an accessible mode, recorded in a tamper-evident ledger, and withdrawal takes effect at once.
• Daily work in one place: timetable, SMART goals, session logging, progress, family reports, leave and cover, reminders and a team calendar.
• Privacy by design: encryption at rest, no trackers or third parties, an access log, and erasure with a deletion receipt.
• Prepared for review: draft privacy notice, consent forms, risk assessment and questions for a lawyer, plus automated tests.

1. Register & consent
2. Assessment
3. Refer to therapy
4. Allot
5. Sessions & reports
6. Supervisor review

Results

• Delivered: a working, installable web app with a full demo clinic, live online and open source.

Read the full case study