I See Data People

The episode packs real technical content into 14 minutes - synthetic controls for news event impact, the reactionary vs. alpha-signal framework, and LLM-based narrative projection are substantive. However, the value is heavily siloed in quant finance and many ideas are explained at a conceptual level without enough follow-through for a practitioner to act on.

“we are looking into using modeling techniques called synthetic controls and synthetic intervention, or a more recent model from Google's causal impact machine learning models to quantify near-term event impact on stocks' relative performance post-news event”

“access return by definition reduced the correlation across the stocks, even though the stocks can be from the same sectors”

The idea of using causal-impact models for single-stock news events in a market-making context is genuinely non-standard, and predicting future narratives via LLMs is an interesting framing. But the guest's own 'controversial' view - that fundamental data matters in short-term trading - is by her own admission now mainstream, blunting the originality claim.

“news sentiment was once considered as an alternative data is now becoming more mainstream. And my novel idea was some controversial view I had actually from a decade ago. is becoming more palatable today”

“we don't construct portfolios and allocate risks like buy-side investors. But rather, our risk is predominantly decided by our client order flows”

Judith Gu is a legitimate practitioner who built a quant trading business at Scotiabank after VP-level roles at Goldman Sachs and Citadel - she has genuinely done the thing at scale. The short runtime limits how much depth is extracted from her credentials, but the credential-to-content ratio is solid.

“my team runs the equities-quant trading on a market-making desk. At Scotiabank, we cover both Canada and U.S.”

“Over five years ago, when we started building this quant trading business, we never had to load any raw data from vendor”

The episode names specific techniques (synthetic controls, Google's causal impact model, vector databases) and specific data artifacts (corporate action adjustments, security master, NBBO), which is creditable. However, there are zero performance figures, no concrete outcomes from implementing any approach, and no timelines beyond 'over five years ago' - the specificity is terminological rather than evidential.

“Corporate actions like cash, stock dividend, stock splits happens every day given the stock universe. And historical prices and returns need to be adjusted to get the correct returns”

“The collaboration between cloud technology and other database, more performance-driven database like Vector Database, can bring down the technology barriers in a material way in the future”

The hosts use a rigid, pre-planned question template ('what data do you wish you had?', 'most powerful insight?', 'most controversial view?', 'where is data going in five years?') with no meaningful follow-up when interesting threads appear, and they respond to substantive answers with flattery rather than probing. The interview functions as a structured promotional feature, not a real intellectual conversation.

“Sounds like your controversial opinions are pretty prescient about the future”

“Love this. And you know quantitative trading often requires also low latency data solutions”