Recent work is most useful when it explains the problem, the change and what was actually verified. These notes cover selected work by Shahar Dam Ari during September 2026, with the delivery stage made explicit.
1. MyRentCar: a growth workflow grounded in evidence
Production fix verified · Experiment outcome inconclusive
The work on MyRentCar connected search performance, booking-funnel observations and bilingual content reviews to a controlled publication process. Each proposed change had to answer a specific user need and pass checks for sources, language parity and the booking experience.
During September, a cross-language navigation defect was corrected and the production pages were checked against the reviewed release. Mobile booking reviews also checked whether pickup and date controls and the supplier's Search button were visible on a short phone screen.
The workflow included registered query-and-page experiments with defined observation windows. The airport-versus-city pickup experiment closed as inconclusive: the sample was small, there were no observed search clicks in the registered windows, and a navigation change complicated interpretation. We did not turn a position change into a claim of business growth.
The lesson: a useful growth system can recommend a fix, a new experiment or no change. A booking-page view, widget interaction or provider form event must remain distinct from a completed rental.
2. An AI lab for infrastructure and operations teams
Workshop and demo kit prepared · Local exercises tested
We prepared a Hebrew AI workshop and a supporting demo kit for an infrastructure audience. The material connects AI use to familiar operational tasks: packet-capture analysis, DNS investigation, Bash script review and drafting an incident runbook.
The kit uses synthetic data and a local lab. Additional exercises cover suspicious traffic and access patterns, a controlled HTTP service failure and prompt injection. Presenter notes and a step-by-step runbook support repeatable practice rather than relying on a single successful demonstration.
The preparation included local tests for the scripts and lab data. The runbook is available in English with Hebrew prompts, alongside the Hebrew version. A live Copilot rehearsal still depends on signing in with the approved presentation account; preparation and local tests are not presented as proof that the workshop has already been delivered.
The lesson: teach teams to inspect evidence, separate observations from hypotheses and review proposed actions before running them. A good exercise should work with safe sample data and have a clear recovery path.
3. Cloud delivery with a clear verification boundary
Application release and migration verified in staging
September application delivery also included changes that required a data migration. The release was checked in staging and followed by independent validation of the migrated state. Completion was tied to the deployed behavior and validation result, rather than the fact that a pull request had merged.
This example is intentionally described without client data or internal system details. It demonstrates an operational approach: identify the target environment, understand the migration's scope, verify the release and then check the resulting data separately.
The lesson: staging evidence supports a staging result. Production readiness and a production rollout remain separate decisions, with their own verification.
What this means for your team
These projects share a practical pattern: define the user problem, build the smallest useful change, test the real behavior and leave a clear record of what is complete and what remains uncertain.
- For a growth workflow, start with trustworthy measurement and a specific customer task.
- For AI adoption, give staff a repeatable exercise, a runbook and safe data.
- For application delivery, pair release checks with operational ownership and independent validation.
TSI Integration supports this work through AI strategy and architecture, hands-on implementation and engineering leadership. Share the workflow or delivery constraint you want to improve, and we can identify a concrete starting point.