Look, I’ll be honest. For years I ran bioreactors the way most of us do: set your temperature, set your pH, watch the DO trace, take a sample every few hours, and hope for the best. You tweak the feed rate because “it feels right.” You crank up the stirrer because the DO is dropping.… Read More: Why I Stopped Trusting My Gut (and Started Trusting My… »
The Gap Between Models and Tools I’ve seen a lot of impressive ML models in biopharma that never get used. Not because the science is wrong, but because the tool doesn’t fit into anyone’s workflow. The model might be published in Nature Methods with beautiful receiver operating characteristic curves, but if a discovery scientist can’t… Read More: Building ML Tools Scientists Will Actually Use »
The Features That Matter After training the developability model on 100+ therapeutic antibodies, I looked at the feature importance rankings. The model had learned to weight certain properties more heavily than others when making predictions. Some of these were obvious. Some were surprising. All of them tell us something about what actually makes antibodies difficult… Read More: What Makes an Antibody Hard to Manufacture? Data-Driven Insights »
Why Trastuzumab Is the Perfect Test Case When I built my antibody developability predictor, I knew I needed to validate it against a molecule where we actually know the manufacturing story. Trastuzumab (Herceptin) was the obvious choice. It’s one of the most successful therapeutic antibodies ever made, with decades of manufacturing data behind it. More… Read More: Case Study: Predicting Trastuzumab Developability »
A bioprocess engineer’s journey into machine learning and why the pharmaceutical industry desperately needs this bridge When I tell people I work in bioprocess engineering, I usually get blank stares. When I explain that I help manufacture proteins in giant tanks for therapeutic use, the response is often: “Oh, like brewing beer?” Not quite. But… Read More: How I Built a Machine Learning Tool to Predict Drug… »
There’s a specific kind of tension that defines working at the intersection of artificial intelligence and regulated discovery science. It’s not technical. It’s not about model architecture or compute budgets. It’s about the gap between what a model can do and what a scientist will trust enough to use AI product leadership in biopharma R&D… Read More: What AI Product Leadership Actually Means in Life Science R&D »
I Built a Game Where You Can Watch AI Think (And It’s Weirdly Satisfying) You know that feeling when you’re trying to learn something new and the teacher just says “trust me, it works”? Yeah, I hate that too. So I built a game that does the opposite. It’s a little AI agent learning to… Read More: AI Brain Demo For Fun »
![Description English: Ethanol near IR spectrum. I took this spectrum using an Ocean Optics near IR (NIR-512) temperature-regulated InGaAs detector spectrometer [1] with IR fiber optic light guide. This is a very rough spectrum and should not be used for any kind of quantitative data whatsoever. I took it by shining the light from a halogen lightbulb through a tiny (~20ml) beaker of liquid ethanol (~2cm liquid optical path) and into the fiber optic of the spectrometer (I subtracted the spectrum of the empty beaker before taking this one). The spectrometer was not really intended to be used this way and it is a very sloppy way to take a spectrum! Nonetheless, based on comparing it to simillarly taken spectra of water and methanol (and other professionally traken nir ethanol spectra), with the exception of the region between about 1400 and 1600nm (this region is saturated by very strong absorbance), I think it likely fairly accurately shows real features of the NIR spectrum of this compound. Date 10 September 2006 (original upload date) Source Transferred from en.wikipedia to Commons by Sevela.p using CommonsHelper. Author The original uploader was Deglr6328 at English Wikipedia.](https://kemal.yaylali.uk/cdn-cgi/image/format=auto,slow-connection-quality=30,onerror=redirect/https://kemal.yaylali.uk/wp-content/uploads/2025/11/Ethanol_near_IR_spectrum.png)
I recently had to audit a production NIR spectroscopy system used for biomass analysis. On paper, it sounded impressive: “state of the art chemometrics,” “advanced multivariate calibration,” “industry ready models.” In reality? It was a textbook example of how to do NIR modeling wrong in ways that quietly transfer risk and liability from the software vendor… Read More: NIR Biomass Analysis Gone Wrong »
The Inflection Point Three-four years ago, I stood in front of production equipment I’d designed using textbook equations and decades-old heuristics. The systems worked—they always had—but I couldn’t shake the feeling that we were leaving performance on the table. That nagging intuition led me down a rabbit hole of machine learning papers, Python tutorials, and… Read More: From Conventional Design to AI-Driven Innovation: A Strategic Journey in… »