A lot of my time is spent reviewing models that didn’t perform as expected. I work closely with analysts to unpack why a volatility forecast went wrong, documenting every assumption and decision point. We treat each failure as an opportunity to improve—sometimes the fix is technical, sometimes it’s in the way we interpret the data. My day is a mix of analysis, documentation, and open discussion.
The highlight is when a model failure leads to a breakthrough—when a mistake uncovers a hidden relationship or prompts a change in process. The most difficult part is getting everyone comfortable with admitting errors. It’s a cultural shift, but it’s essential for technical growth. The best teams are those that share their failures as freely as their successes.
Machine learning has made volatility forecasting more powerful, but also more opaque. When a model fails, we spend extra time tracing through its logic and retraining on new data. Peer review has become critical—we need fresh eyes to spot issues that even the original developer missed. The process is slower, but the models end up stronger.
For those new to the field: don’t hide mistakes. Document them, discuss them, and invite feedback. Most technical growth happens after something fails, not when everything runs smoothly. Stay open to criticism—it’s your best resource.
Earlier in my career, I thought mistakes meant weakness. Now, I see them as the main drivers of progress. The field moves quickly, but only those willing to reflect and adapt will keep up. Peer learning and honest critique are now at the core of my process.
