Research, Engineering, and Data Are Different Crafts
How do research, engineering, and data work differ when all three touch the same public market? This note separates the crafts, using a single question about SPY’s opening auction as the example, for internships and hiring in New York, London, and Hong Kong.
One market, three responsibilities
Research asks whether a stated hypothesis survives tests, and writes the conclusion a person is willing to own. Engineering builds the path that makes a record replayable and keeps a notebook from becoming an order. Data work decides whether a file is the file it claims to be: the clock, the population, the discards, and the acceptance. All three can sit in one room and still be different crafts. An interview that treats them as one generic quantitative job will hire a person for a craft the person did not practice, or will reject a specialist for lacking the other two.
Internships should name the craft. A summer spent on a data check is a success if the check ships. It is not a failed research summer. A summer spent on a research note is not a failed engineering summer. New York, London, and Hong Kong each contain more than one craft. Candidates should be told which one the conversation is for. Live capital decisions are none of the three crafts’ internships. They stay with the firm.
Research owns a hypothesis and its tests. Engineering owns a replayable path. Data work owns whether the file is what it claims. An interview should say which craft it is sampling.
The opening auction, asked three ways
SPY’s opening auction is a public mechanism a candidate in any office can read about. Asked as a data question, it becomes: which rows in a file are auction prints, and what breaks if they are mixed into the continuous session? Asked as an engineering question, it becomes: how would you replay a morning so the auction and the later prints stay distinct, and what do you log when the sequence jumps? Asked as a research question, it becomes: what hypothesis about the open is being stated in advance, and what result would kill it? None of these questions is a request to trade the open.
A candidate who answers all three the same way has not heard the craft. The data answer that never mentions the hypothesis is complete for a data interview. The research answer that never mentions a parser is complete for a research interview. Interviews go wrong when the panel rewards only the answer that sounds like every craft at once. Interns, in particular, should not be punished for failing to be three departments. The internship is where one craft is practiced with a reviewer.
The offices do not specialize the crafts into stereotypes. Engineering in Hong Kong is not a lesser version of engineering in New York, and research in London is not a different method. The trading day arrives at different local times, which changes when a person can watch an open. It does not change whether the open is an auction. Candidates preparing for a city should prepare for the craft named in the interview, and they may mention which of New York, London, and Hong Kong they can work from. They do not need a city-specific theory of markets.
Hire the craft you named
A process that names the craft in the interview and then judges a different one has wasted the candidate’s preparation. The remedy is to write the craft on the conversation before the conversation starts, and to ask the auction question in only one of its three forms unless the role truly spans them. Most roles do not span them on day one. Internships almost never should.
Research, engineering, and data are different crafts that share a market. The SPY open is the shared example. The interview, in New York, London, or Hong Kong, should be willing to say which craft it is sampling.
This piece does not assign a craft to a person, describe openings, or blur internships into live trading. It separates three kinds of work, in New York, London, and Hong Kong. One question about the SPY opening auction, asked three ways, is the example.
