Inside the Quant Researcher, Trader, and Developer's Day
What 10 hours at Citadel, Jane Street, and Two Sigma actually looks like, hour by hour
The job posts get romanticized to death. Recruiters sell the work as math wizardry and screens flashing green. The reality is far more granular, far more tedious in places, and far more lucrative than the outside view suggests.
What follows is what the three core seats actually look like across Citadel Securities, HRT, BNP Paribas AM, RBC Global Asset Management, Jane Street, and the broader systematic landscape, drawn from current practitioners describing their own days.
The titles sound similar from the outside, but the work is not.
The Quant Researcher
The romantic version of QR work has you sitting in a quiet room deriving stochastic differential equations. The real version is closer to a PhD program in physics where the dataset happens to be financial markets.
A current QR at a mid-size systematic fund breaks the time roughly like this. Around 60% of the day gets spent finding patterns and relationships in data, which in practice means scripting and data analysis over some dataset, then using those relationships to predict quantities relevant for the trading model.
Another 20% goes to coding models, debugging existing ones, and pushing production Python. The remaining 20% covers whiteboard sessions with colleagues, reading papers, chasing relevant concepts, and the meeting overhead that any team inherits.
The HFT researcher at an established London firm describes a structurally different day that still rhymes. The math is largely ‘done’ on core problems. Working models for central questions already exist. Iterative improvements happen constantly. Full rewrites are rare. The actual work falls into three buckets.
The first is signal generation and model improvement through better machine learning, deeper microstructure understanding, and integration of new data feeds. The second is application of those models to live trading, because even a perfect signal does not trade itself.
As one HFT researcher put it, actually trading involves managing a whole load of different concerns and trade offs. This is essentially one big, continuously changing, optimisation problem. The third is opportunity identification across new exchanges, new countries, lower fees through different LP schemes, and tier moves. The third is the least interesting work, in his own words, but it has to be done, because there’s no point spending weeks on a fancy model if we could just start trading more symbols next week and make larger gains like that.
Sample projects from the last year at one HFT researcher’s desk include modeling how aggressive trades impact the market and incorporating those effects into backtest, integrating new information streams by building fresh signals on top of data piped in by someone else, and studying T+1 adverse selection in specific symbols and what that implies for position management.
Most of the production code is Python for building and training models. All the actual trading is done on FPGAs which is handled entirely by Devs. Collaboration with engineers runs deep, particularly when getting a neural network onto hardware, which the same researcher described as quite delicate.
Project length usually runs a few weeks plus another couple to get it into production. A decent share of projects get killed within a week because they turn out unworkable. The kill discipline is aggressive. As the same researcher put it, if I don’t have a good angle within a week it gets put in the backlog. Most researchers keep two projects active at once to minimize dead time.
One QR was honest about motivation: it becomes a similar arbitrary mix of social prestige and curiosity like it is in academia. The status games are real and invisible to outsiders. People are extremely susceptible to status games of their immediate peers that the general public do not know or care about. Chess players chase GM titles, academics chase citations, and quants chase pod allocations and PnL at the top firms. The games look meaningless from outside the group and feel like everything inside it.
The Quant Trader at Citadel Securities Miami
Here is a verified schedule from a Citadel Securities FX quant researcher based in Miami who joined the firm in 2021 after a PhD in physics at Cambridge.
He wakes between 7 and 8 AM, lives a five-minute walk from the office, and tries to keep things civilized with a morning jog or cycle, sometimes a morning swim at the beach ten minutes away. Because my hours aren’t dictated by the markets, I can ease into the day more than those in some other roles in finance.
8:30 AM is office arrival. He checks overnight PnL on the running strategies, prioritizes pressing issues including any bugs raised, and reviews results of the compute jobs that completed overnight.
9:30 AM is collaboration with European colleagues. FX is a 24-hour market. I’m only working for a portion of that period, so it’s good to understand what my colleagues in London and elsewhere have been seeing.
10:30 AM is the direct reports meeting. He started managing people just a couple of years after joining Citadel Securities, which doesn’t happen at many firms but isn’t at all unusual here.
11 AM to 12:30 PM is strategy refinement, model building, and deployment of systematic trading algorithms. 95% of the time, our strategies don’t require human input. They’re fully systematic. The job is monitoring, improving, and extending those strategies to new products. The feedback loop is fast because work hits the market quickly.
12:30 PM is catered lunch with bay views, sometimes at the desk if focus is high.
1 PM to 3 PM is more strategy refinement built around post-trade analysis on what worked and what did not.
3 PM is a cross-asset meeting with other QRs to understand how they manage risk and opportunity outside FX, with an eye on whether anything is transferable.
3:30 PM to 6 PM is strategy work, sometimes new strategies, sometimes applying existing strategies to different opportunity sets. He works closely with traders on risks that are difficult to model, like upcoming economic events, and with engineers whose infrastructure powers the trading.
He typically leaves around 6 PM, sometimes later. 7 PM is the track club in Miami. 9 PM is occasional evening work, because I enjoy working late at night because there aren’t any interruptions and I can really focus. Midnight is sleep.
Roughly 9.5 hours in-office plus optional evening work. The cadence is recognizable across systematic firms regardless of asset class.
The Quant Trader at RBC Global Asset Management
Different firm, different asset class, sharper market-driven punctuation, same general rhythm.
7 to 9:30 AM starts with the trader plugged in by 7 AM Eastern. First, I check my live orders from Europe, which are in mid-trading session at this time, then I check overnight fills from APAC. He confirms comfort with the current strategy or adjusts as needed. Pre-North America open includes reading research reports, taking broker calls, implementing trading strategies on new and multi-day equity orders, and sharing highlights with Quant PMs.
9:30 AM to noon is execution. At 9:30 am, the North America market opens and dominates my attention. The primary responsibility is executing orders while keeping market impact minimal, using algorithmic and block trading. He also analyzes and presents trade opportunities to PMs based on the activity of other market participants.
Noon to 4 PM is monitoring with no real break. No lunch time for traders. You can’t afford to be out of the desk if something happens. Reaction time has direct PnL implications. In just seconds can really cause a multi-million loss or gain, so you have to be connected at all times. He brings food from home or grabs something nearby and eats at the desk. Every moment away requires a backup trader on coverage.
4 PM is close. He books trades, catches up on email, and uses the post-close window for side projects, backtesting new strategies, and running simulations.
The shape is consistent across Citadel Securities and RBC. Pre-market is preparation. Open is execution intensity. Midday is monitoring with no real break. Close is reconciliation, reporting, and the only quiet window for personal research.
The Quant Developer
Quant dev gets the least glamorous press and arguably the most operational leverage per line of code. Here is what the seat looks like at a Mayfair hedge fund.
6 AM is breakfast and an email check to confirm overnight cron jobs ran successfully, covering financial data downloads and internal report uploads.
7 AM is commute reading. Algorithmic trading textbooks, market access literature, FT.
8 AM is the start of a mix of diagnosis and repair of abnormalities in the infrastructure that we had developed, as well as development of new requested features. Remaining automated data tasks get verified. Any incomplete jobs get fixed immediately so they do not happen again. RSS feeds for trading ideas and developer tooling get scanned.
9 AM is a brief catchup with the lead QR on data requests, infrastructure requests, and US market state.
10 AM is the first real fire. A late-running cron script failed because of an undocumented external API change. He modifies unit tests, pushes to staging, and deploys to production. Since our code has good test coverage it is not a problem to be continuously deploying.
Noon is lunch off-desk. I very rarely have a “working lunch” at the desk since I don’t like to eat and code. He reads trading strategy material in the park, taking notes.
1 PM is US market prep. The Portfolio and Order Management System pings the brokerage API every ten minutes to pull the current state of the portfolio. The diff against the ideal portfolio gets sent to the brokerage. MOO orders execute on the open.
2 PM is new data sources, writing Python download scripts for a new fundamental data API, scheduled via cron.
3:30 PM is the spec for a spike checker that emails the lead QR and the dev if any end-of-day price moves more than 20% from the previous daily bar, letting the team handle corporate actions before downstream models break.
5 PM is the management meeting. A traffic light system covers issue severity. Recent fund performance gets reviewed against backtests. Operational issues, new data sources, and new strategy ideas get worked through. New automation tasks get prioritized.
6 PM is home. Reading continues into the evening.
Strip away the romance and quant dev work is not fundamentally different than any software dev role. A large portion is shuffling data either within a single application or process or between processes. The core logic of many applications is fairly simple code-wise. Most of the work is getting all the inputs into the right place.
Three traits separate the quant devs who do well from the ones who do not. First, extreme attention to detail, because the business is much more winner-take-all than other industries so every bit counts. Second, the ability to handle vague requirements and iterate quickly, since trading has been a largely intuitive domain for over a century. Algorithmic trading is roughly 20 years old, and the first decade ran on simplistic algorithms that were making 100K per day in 2005.
Many experienced people on the trading floor are not particularly analytical, which makes translating their expertise into code an actual skill. Third, persistence with a willingness to fail. Most ideas fail. The industry attracts the smartest and most driven people in the world. As one senior dev put it, in any given room, you’re probably not the smartest, even if you’re really fucking smart. However, regardless of talent, without effort you’ll never find success.
Quant dev specialization is largely a question of scale. A 50-dev shop has limited specialization, mostly split by asset class. A Citadel or Jane Street with over a thousand developers has hyper-specialized teams for latency, networking, time-series storage, internal trader UI dashboards, and execution routing optimization at the SIMD level.
The Hours Question
One systematic hedge fund QR ran the numbers on himself for a year. The average came out to 10.7 hours per day, or 53.5 hours per week, with a range from 7 hours minimum to 13 hours maximum and most days landing between 10 and 12. Lunch averaged 45 minutes. Commute was 25 to 30 minutes each way.
He described himself as average, not a superstar. Hours were reasonable compared to investment banking and longer than most engineering jobs. Flexibility was high for late starts, early leaves, and occasional WFH. Out-of-hours work existed but was rare. Compensation was quite high, and has attractive distributional qualities, meaning the minimum was still solid and the right tail was long. Some grind work was unavoidable, including data cleaning and reporting, anything that was not pure research or production trading code.
A risk management quant at an O&G major reports a standard 9-to-5 in industry and longer in consulting. The risk seat trades pay ceiling for sane hours. Insurance and bank risk management cluster around the same shape, with CCAR stress testing seasons spiking hours significantly at US bank holding companies.
The HFT QR at the established London firm captured the time-allocation problem precisely: honestly correctly prioritising work is the difficult part, not idea generation.
Compensation Reality
For full access to daily quant articles (4$ a month):
