What Mercor's finance AI interview covers, the questions its finance and risk listings point to, and a worked expected-loss example to practise.
Mercor does not publish a separate interview format for finance or financial risk roles. Its finance talent-network listings send applicants through a short AI interview of about 20 minutes after a resume upload, and Mercor's own help pages say some of its AI interviews are technical and require deep discussion of your prior experience. So the Mercor finance interview questions worth preparing are the ones that test the work the listings describe: building a model or analysis yourself, checking whether someone else's numbers hold up, and explaining your judgement clearly enough that it could become a grading rubric.
Below are what Mercor documents, what its finance and risk listings describe, six questions with illustrative answers, and one risk calculation we ran in Python; for mechanics shared by every domain, see our guide to how the Mercor AI interview works.
What Mercor documents about the interview
Mercor's AI interview page says an AI interviewer asks role-specific questions, produces a transcript and evaluates your performance, usually in about 20 minutes, and that a result carries over to other applications using the same interview. It also tells you to have pen, paper and a calculator ready if relevant, warns that screen sharing may be required and that only full-screen sharing is supported, and notes that long pauses can make the AI interject. You can retake an official interview up to three times across all applications, but the waiting room asks you to keep retakes for technical problems, and hiring managers can see your retake count. A free practice interview does not affect future opportunities.
The Finance Expert and Investment Banking Expert listings add the sequence: apply with your resume, confirm where you work from, take the AI interview of about 20 minutes, and have your background verified before you join the pool. They say plainly that they do not accept or reject candidates, and Mercor instead invites people from the pool to a project listing, which can happen within a week or after several months.
Some projects then add assessments, which Mercor says usually allow one submission and may be timed, or a single paid work trial per person; our page on the Mercor skills assessment and work trials covers both stages.
We looked for dated candidate reports specific to Mercor's finance interview and could not find any we were able to open and verify, so this page does not repeat second-hand claims about question counts or follow-up patterns.
What the finance and risk listings say the work is
The listings are the best public evidence of the work the interview leads to, so they are a reasonable guide to what to prepare. The Finance Expert network, which covers accounting and auditing, corporate finance, investment banking, asset and wealth management, and insurance and actuarial science, describes three kinds of task. You write a financial problem from work you have done, with filings, assumptions and a reference answer; you grade model-built financial work on whether it ties, whether the accounting treatment is right and whether valuation assumptions survive a sensitivity check; and you flag output that looks clean but would not pass review.
In the listings we found on Mercor's job board on 25 October 2026, risk appears inside several networks rather than as one dedicated financial risk role. The Compliance & Risk Specialist network asks for experience in regulatory compliance frameworks, risk assessment and mitigation, and internal controls and audit. The Asset & Wealth Management network describes writing an investment case with its constraints and risk budget, and flagging recommendations that would breach a real mandate given concentration limits or drawdown tolerance. A separate Finance Domain Expert role, a hybrid W-2 position in the Bay Area through Cincinnatus LLC, names quantitative finance and risk management as one qualifying discipline and lists the CFA, CPA and FRM among preferred credentials.
Our reading of these listings is that the interview is checking two things. The first is depth: the Investment Banking listing says the firm name matters less than whether you built the models and wrote the materials yourself, recently enough to explain the judgement calls behind them. The second is whether you can say why a piece of work is wrong, in writing or out loud, since the listed tasks involve writing reference answers, grading work and flagging problems.
For how Mercor's expert interview works across fields, see our guide to the Mercor Domain Expert Interview.
The questions worth preparing
The answers below are short illustrations of the shape we think works, not scripts, so replace every detail with your own experience.
Walk me through an analysis you built yourself
This tests whether you did the work yourself, and a strong answer names the decision the analysis supported, the inputs you chose, the one assumption that drove the result, and what you would change today. Illustrative answer: "I built the credit model for a refinancing of a mid-sized distributor; leverage looked fine, but the covenant headroom depended on a working-capital release I did not trust, so I re-ran it without the release and we tightened the covenant package."
How would you check whether a model ties?
The Finance Expert listing names this directly. Illustrative answer: "I check that the balance sheet balances in every period, that ending cash on the cash flow statement equals cash on the balance sheet, that net income flows into retained earnings net of dividends, and that debt schedules roll forward from opening balance to closing balance before I look at any valuation output."
This valuation looks reasonable. How do you test it?
This tests whether you look for the assumption that carries the answer. Illustrative answer: "I find the inputs that move the value most, usually the terminal growth rate, the discount rate and the margin path, and flex each one over a range I can defend. If a small change in one input moves the value more than the deal can absorb, I say so in the review, even when the base case is fine."
What does a 99% one-day VaR of $1.4 million tell you, and what does it not?
This core risk-measure question is easy to answer too confidently. Illustrative answer: "If the model's assumptions hold, losses should exceed $1.4 million on about one trading day in a hundred, but VaR says nothing about how large the loss is on that day, which is why I would also look at expected shortfall and stress tests." The worked example below shows where that number comes from.
How do you assess whether a control works?
This fits the Compliance & Risk network's internal-controls requirement. Illustrative answer: "I test design first, asking whether the control would prevent or detect the risk if it ran exactly as written, and then operating effectiveness, which means sampling instances across the period and checking the evidence that it actually ran, by the right person, on time."
How would you grade an answer that reads well but is wrong?
The listings describe writing rubrics and grading model output, so expect a version of this. Illustrative answer: "I write the criteria before I read the answer: the correct figure, the method that gets there, and the errors that should cost marks. Then I note which criterion failed and why, so another grader would reach the same score."
Worked example: expected loss on a small loan book
The Basel Committee's explanatory note on the IRB risk-weight functions writes expected loss as EL = PD × EAD × LGD, where PD is the probability of default, EAD is the exposure at default, and LGD is the share of that exposure lost if the borrower defaults. The note defines EAD as drawn amounts plus likely future drawdowns of undrawn lines, which is easy to overlook.
The numbers below are invented for practice, not taken from Mercor or any real portfolio, and the output is what the script printed when we ran it in Python 3.
from statistics import NormalDist
from math import sqrt
# name, drawn, undrawn, share of undrawn assumed drawn at default, PD, LGD
loans = [
("A secured term loan", 10_000_000, 0, 0.0, 0.02, 0.40),
("B revolver", 5_000_000, 5_000_000, 0.5, 0.04, 0.45),
("C unsecured loan", 3_000_000, 0, 0.0, 0.08, 0.75),
]
total_ead = total_el = 0
for name, drawn, undrawn, ccf, pd, lgd in loans:
ead = drawn + ccf * undrawn
el = pd * lgd * ead
total_ead += ead; total_el += el
print(f"{name:22s} EAD {ead:>12,.0f} EL {el:>10,.0f}")
print(f"Total EAD {total_ead:,.0f} Total EL {total_el:,.0f} EL/EAD {total_el/total_ead:.2%}")
print(f"Revolver EL on drawn only {0.04*0.45*5_000_000:,.0f}")
print(f"Loan A EL at LGD 60%: {0.02*0.60*10_000_000:,.0f}")
# Parametric (normal) VaR and expected shortfall on a separate $50m portfolio
V, sigma = 50_000_000, 0.012
z = NormalDist().inv_cdf(0.99)
var1 = z * sigma * V
print(f"z {z:.3f} 1-day VaR {var1:,.0f} 20-day VaR {var1*sqrt(20):,.0f} "
f"1-day ES {sigma*V*NormalDist().pdf(z)/0.01:,.0f}")A secured term loan EAD 10,000,000 EL 80,000
B revolver EAD 7,500,000 EL 135,000
C unsecured loan EAD 3,000,000 EL 180,000
Total EAD 20,500,000 Total EL 395,000 EL/EAD 1.93%
Revolver EL on drawn only 90,000
Loan A EL at LGD 60%: 120,000
z 2.326 1-day VaR 1,395,809 20-day VaR 6,242,246 1-day ES 1,599,129The book's expected loss is $395,000, or 1.93% of exposure. The point to say out loud is the revolver: using only the $5 million drawn cuts its expected loss to $90,000 and understates the book by $45,000, which is why the Basel definition of EAD counts likely drawdowns of undrawn lines. If Loan A's collateral is worth less than assumed and LGD rises from 40% to 60%, that loan's expected loss rises from $80,000 to $120,000, so a reviewer should ask how LGD was estimated before accepting the total.
The last line is the VaR from the question above, for a $50 million portfolio with 1.2% daily volatility, assuming normally distributed returns. We scaled it to 20 days with the square root of 20, which only holds if daily returns are independent. We chose 20 days because the SEC's Rule 18f-4 for registered funds requires a VaR model with a 99% confidence level and a 20-trading-day horizon. The expected shortfall of about $1.6 million is the average loss on days worse than the VaR, which is the answer to what VaR does not tell you.
How to prepare in the week before
Mercor's help pages point to a short, practical plan, and our page on how to prepare for and pass the Mercor AI interview covers the setup checks in more detail.
- Pick two or three pieces of work you built yourself, and for each one write down the decision it supported, the key assumption and the check that caught a problem.
- Practise explaining one calculation aloud, such as the expected-loss example above, in under two minutes without notes, because long silences can make the AI interviewer interject.
- Take Mercor's free practice interview first, since it does not affect future opportunities and your official retakes are visible to hiring managers.
- Have pen, paper and a calculator on the desk, use Chrome, and test your microphone and camera in the waiting room before you start.
- Write your own answers, because Mercor's AI interview help page says to follow its LLM usage policy, which allows grammar and wording checks but not using large language models to compose your responses, and its assessments page says the same for assessments unless stated otherwise.
Practising with Verve AI
Because Mercor's policy does not allow large language models to compose your answers, practise before the interview instead. Verve AI's mock interview has an AI recruiter ask you questions one at a time. It offers a Finance question bank and a Custom Questions option that lets you practise questions from your own Question Bank, so you can add the six questions on this page and rehearse them. After the session, you receive a detailed performance report with personalized feedback and scoring.
FAQ
Does Mercor have a separate interview for financial risk roles?
Mercor does not document one, and its help pages describe an AI interview with role-specific questions. Risk work appears inside several listings, including Compliance & Risk Specialist, Asset & Wealth Management and Finance Domain Expert, rather than as one dedicated role. Prepare for questions about your own risk work and about how you would check someone else's.
Can I use a calculator or AI tools during the interview?
Mercor's help page tells you to have pen, paper and a calculator available if relevant. Mercor's AI interview help page tells you to follow its LLM usage policy, which allows grammar and wording checks but not using large language models to compose your responses. Assessments add their own restrictions, and some disable copy and paste.
What happens after the finance interview?
The Finance Expert, Investment Banking Expert and Asset & Wealth Management listings say they do not accept or reject candidates. Mercor verifies your background, adds you to the pool, and invites you to a specific project listing when one matches your experience, which can take a week or several months. Some projects then use assessments or a single paid work trial before an offer.
Sources
- AI interview length, role-specific questions, calculator, screen sharing, retakes, practice interview, data use: https://talent.docs.mercor.com/support/ai-interview
- Assessments: single submission, timed, copy and paste disabled, AI use prohibited: https://talent.docs.mercor.com/how-to/assessments
- Work trials: one paid trial per person, reviewed by the project lead: https://talent.docs.mercor.com/how-to/work-trial
- LLM usage policy: https://talent.docs.mercor.com/policies/llm-usage
- Finance Expert network: task types, fields covered, 20-minute interview, no accept/reject: https://work.mercor.com/jobs/list_AAABoKOQo_tcpjmZ2_xLb7qN/finance-expert
- Investment Banking Expert network: firm name versus building the models yourself, interview and background verification: https://work.mercor.com/jobs/list_AAABoKMhfQur1DCRx9tAD7Jq/investment-banking-expert
- Compliance & Risk Specialist network requirements: https://work.mercor.com/jobs/list_AAABnJ0Tj07Vgzf7imVMBZB6/compliance-risk-specialist-talent-network
- Asset & Wealth Management network: risk budget, concentration limits, drawdown tolerance: https://work.mercor.com/jobs/list_AAABoKMi5yd-bKOyjYtOUZox/asset-wealth-management-expert
- Finance Domain Expert (Cincinnatus LLC, Bay Area, W-2): risk management discipline, CFA/CPA/FRM: https://work.mercor.com/jobs/list_AAABn_Iz3TRXq5o1-J5Gdot8/finance-domain-expert-ai-training-evaluation
- Expected loss formula and EAD definition (Basel Committee, IRB explanatory note): https://www.bis.org/publications/explanatory-note-basel-ii-irb-risk-weight-functions.pdf
- SEC Rule 18f-4 VaR model: 99% confidence, 20 trading days: https://www.sec.gov/resources-small-businesses/small-business-compliance-guides/use-derivatives-registered-investment-companies-business-development-companies-small-entity
- Verve AI mock interview: AI recruiter, Finance and Custom Questions banks, performance report with feedback and scoring: https://docs.vervecopilot.com/features/mock-interview
Ryan
CEO of Verve AI










