<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Nassim Louissi — Research Engineer</title><description>Research Engineer at Quinze-Vingts Hospital in Paris. Builder of CorneaForge / OphtaFlow AI, working across deployed ophthalmic systems, representation learning, and GPU performance.</description><link>https://nassimlouissi.com/</link><item><title>Reading the geometry of AS-OCT pretraining</title><link>https://nassimlouissi.com/blog/reading-pretraining-diagnostics/</link><guid isPermaLink="true">https://nassimlouissi.com/blog/reading-pretraining-diagnostics/</guid><description>Measured representation changes across five AS-OCT pretraining runs, and the profiling work that cut two-GPU JEPA update time by 14.2%.</description><pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Learning without labels: what MAE and JEPA actually predict</title><link>https://nassimlouissi.com/blog/learning-without-labels/</link><guid isPermaLink="true">https://nassimlouissi.com/blog/learning-without-labels/</guid><description>Follow one image from pixels to representations, calculate a training loss, and see why agreement alone can teach a model nothing useful.</description><pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Five OCT pretraining recipes, and what a fair comparison means</title><link>https://nassimlouissi.com/blog/five-oct-pretraining-recipes/</link><guid isPermaLink="true">https://nassimlouissi.com/blog/five-oct-pretraining-recipes/</guid><description>The five AS-OCT recipes I implemented on 3.43 million images: their inputs, objectives, shared controls, and compute tradeoffs.</description><pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate></item><item><title>From a device export to a research system</title><link>https://nassimlouissi.com/blog/from-device-export-to-research-system/</link><guid isPermaLink="true">https://nassimlouissi.com/blog/from-device-export-to-research-system/</guid><description>How I built and operate CorneaForge / OphtaFlow AI: device ingestion, scientific processing, examination review, and model serving on hospital infrastructure.</description><pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate></item><item><title>From measured points to Zernike coefficients</title><link>https://nassimlouissi.com/blog/from-corneal-points-to-zernike/</link><guid isPermaLink="true">https://nassimlouissi.com/blog/from-corneal-points-to-zernike/</guid><description>How I turn native corneal samples into maps and Zernike descriptors in CorneaForge, with explicit reference surfaces, coordinates and units.</description><pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Cache the geometry, then change the values</title><link>https://nassimlouissi.com/blog/caching-geometry/</link><guid isPermaLink="true">https://nassimlouissi.com/blog/caching-geometry/</guid><description>How I implemented reusable interpolation operators in CorneaForge, including missing-data cache keys and a measured memory budget.</description><pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate></item></channel></rss>