Ved at klikke på "Accepter" accepterer du lagring af cookies på din enhed for at forbedre webstedsnavigation, analysere webstedsanvendelse og hjælpe med vores markedsføringsindsats. Se vores privatlivspolitik, vilkår for brug og tredjeparts tjenester for mere information.
Implemented Home Screen for all the tenants: Lufthansa, Swiss, Brussels,
Austria which uses millions of users. Winner of 'Battle of the Airline Apps’
2024.
Decreased app size 40% by optimizing modularization, linking, file
format.
Developed a Server Driven UI (SDUI) new way to update and manage the
user interface on iOS apps for Lufthansa, Austrian, Brussels, and Swiss
airlines using SwiftUI.
Worked on changing a Python machine learning model (Stable Diffusion)
to be used with CoreML. Wrote an article about how I did it: https://shorturl.at/hxJ02."
Explored how to predict what users might do next using CoreML. Shared
Lead platform enablement for 4 cross-functional feature teams (~20
engineers), focusing on accelerating development cycles and engineering effectiveness.
Improved 30x build & test time on CI (from 25 minutes to 49 seconds),
due to platform optimization (arm64), selective testing and caching
modules.
Fine-Tuned Model.
Built for prompts. Only prompts.
Sapa does not use a general-purpose LLM. The model was fine-tuned on prompt engineering patterns, security anti-patterns, and token efficiency techniques. It understands what makes a prompt good or dangerous in ways generic models miss.
Token Optimizer. Same result. Fewer tokens. Lower cost.
The model restructures your prompt - removes redundancy, clarifies intent, improves context layout - so the receiving LLM needs fewer tokens. Average reduction is 38%.
100% Private. Your data never leaves your machine.
Sapa runs a fine-tuned model entirely inside your Mac memory. Your prompts are processed locally and discarded immediately - never sent to any server.
Security Scan. Catch what you almost sent.
Before you hit send, Sapa detects credential leaks, PII exposure, API keys, connection strings, and injection patterns. Plain-English explanations tell you the risk, not just that one exists.
