What Your First Two Years as a Quant Researcher Actually Look Like at Citadel, Jane Street, and Two Sigma
The offer letter is the easy part. Here is what the seat demands before year two decides your career.
Jane Street lists $300,000 as the base salary on its public New York quantitative researcher posting, before any bonus. Two Sigma‘s campus-hire posting lists a base band of $200,000 to $220,000. Citadel reaches a first-year total near $336,000 at the entry level according to aggregated offer data.
Those figures get screenshotted and passed around every recruiting cycle. They tell you very little about the job itself.
For daily Quant articles and market breakdowns:
An offer letter is a price the firm sets for a person who has not yet produced a single piece of production research. What decides whether you keep the seat happens entirely after you sign.
Here is what the first two years contain at the three firms most students rank first.
The number, decoded
The headline figure hides three parts, and they behave very differently once you are inside.
Base salary is the stable floor. At the top firms it sits between $150,000 and $300,000, and it barely moves across an entire career.
Jane Street and Five Rings both anchor at a flat $300,000 base for New York research hires, a figure confirmed across public postings and H1B filings. Two Sigma campus hires start lower, at $200,000 to $220,000, with the balance made up through bonus.
Bonus is where the dispersion lives. First-year total compensation at a strong research seat runs $250,000 to $500,000 once a signing bonus and, in many cases, a first-year guarantee get added on.
That guarantee matters more than most candidates realise. A large part of your year-one total is one-time money that will not repeat, which means your year-two figure can come in lower even when the underlying work has improved.
Deferred compensation is the part nobody explains in the interview. A meaningful share of the bonus vests over several years. Firms structure it this way deliberately, to raise the cost of leaving for a competitor before the money has been paid out.
The rough first-year totals, combining base, signing, guarantee and bonus:
Jane Street, New York: $400,000 to $700,000, inflated by one-time components
Citadel: entry-level total near $336,000, with the researcher-population median closer to $396,000
Two Sigma, New York: $275,000 to $425,000
London packages run lower in nominal terms, typically £195,000 to £285,000 at Two Sigma and £250,000 to £500,000 at Jane Street. New York pays a premium of roughly 10% to 20% at the same level.
The gap between these firms at the junior level is smaller than the offer screenshots suggest. It widens sharply by year five, and that widening comes directly from what happens during the two years described below.
Months one to six: the research has not started yet
The early months go to absorbing infrastructure. Alpha generation comes later.
You inherit a backtesting stack that someone spent close to a decade building. Your first commits are small features and fixes to that library, reviewed line by line by a senior researcher.
A typical early pull request adds a feature to the team’s backtest framework and comes back with comments about unnecessary copies of large DataFrames before it gets approved. This is normal. The point of the first months is fluency in the firm’s tooling. Output comes after that.
For full access to daily quant articles (12.5$ a month):
50% off paid sub
Once you are ramped, the daily split becomes recognisable:
Reading and idea generation, 20% to 30%. Papers, market data, and conversations with traders and senior researchers.
Coding, 30% to 40%. Signal calculation, backtest framework work, statistical tests. Python dominates, with C++ where latency matters.
Analysis and write-up, 30% to 40%. Reading backtest output, debugging results that look wrong, documenting findings in the team research log.
Meetings, 5% to 15%.
The research log is the part academics tend to underestimate. Six months after you run an experiment, someone will need to reconstruct why you tried it and what you concluded. Undocumented work is treated as work that never happened.
The mental adjustment is the hard part. Most people arrive from a PhD where finishing a project means publishing a result.
Here, the majority of your projects end with the finding that the idea fails once transaction costs are applied. Being comfortable with that outcome is the real entry requirement, and no interview round measures it.
If you are working toward one of these seats, or you are already in the field and considering a move across the trading-technology and HFT engineering stack, Autonomai places people into exactly these roles. The team understands how these firms evaluate junior research talent, and what separates a candidate who clears year two from one who does not. Start at autonomai.io.
Your first real signal, and why it probably dies
Once you own a research question, the work turns into a series of attempts to disprove your own idea before the firm commits capital to it.
Say you are testing options-implied skew as a predictor of equity returns. The preliminary analysis looks interesting enough to justify a full backtest.
Then the signal has to survive a series of checks:
Turnover. Does it trade so often that costs remove the edge?
Transaction-cost sensitivity. Does a realistic cost assumption eliminate the P&L?
Correlation with existing signals. Is the desk already capturing this through something it already runs?
Out-of-sample behaviour. Does it hold on data you did not touch while building it?
Regime stability and data leakage. Did a lookahead bug create the result?
Most signals fail at least one of these. A version that works on the full universe stops working when you restrict it to large caps. Another survives the turnover and correlation checks, then proves too sensitive to cost assumptions to deploy.
A researcher can spend three weeks on a signal that ends in a research-log entry reading “does not survive costs,” and that still counts as a well-executed three weeks.
The reason the checks are this strict is capital. A single production signal can be worth tens of millions to the firm over its life, and a bad deploy carries a real drawdown cost. So the measure that matters for you is your hit rate on signals that reach production.
Where the three firms stop looking alike
The daily mechanics look similar across systematic firms. The environment around them does not, and by month twelve those differences affect your trajectory more than the pay gap does.
Citadel
Citadel runs a multi-manager structure. Research feeds portfolio managers who own P&L. Capital allocation across seats is discretionary and less even than at peers. The long-term prize is the PM track, where a book of your own can pay eight figures in a strong year.
