← intercontinental exchange Interview Insights

intercontinental exchange·Data Scientist·Technical Phone Screen·Senior

Senior
May 2026

Summary

Technical phone screen for a Data Scientist role at ICE, heavy on quantitative finance. The whole thing centered on implied volatility and Black-Scholes, which I wasn't expecting to go that deep on for a data science position.

Questions Asked (3)

Q1

Explain how Newton-Raphson works for finding implied volatility, and walk through the specific conditions where it breaks down, including issues with the objective function, its derivative, step sizes, and edge cases.

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

I knew the basic mechanics but stumbled when they pushed on failure modes.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by explaining the Newton-Raphson method for implied volatility, emphasizing the objective function (difference between market and model prices) and its derivative (vega). Then systematically discuss the conditions where it fails, covering issues with the objective function, derivative, step sizes, and edge cases, and briefly mention practical remedies.

Pro tip: Highlight that while Newton-Raphson is fast, its convergence is not guaranteed; in production, it's often combined with a fallback like bisection or a robust initial guess to handle failures gracefully.

1. Explain the Newton-Raphson method for implied volatility

Describe the iterative formula: σ_{n+1} = σ_n - (BS(σ_n) - MarketPrice) / Vega(σ_n), where BS is the Black-Scholes price and Vega is the derivative with respect to volatility.

2. Discuss issues with the objective function

Mention that the objective function (price difference) may not be monotonic or well-behaved, especially for deep in/out-of-the-money options or near expiration, leading to multiple roots or no root.

3. Address derivative (Vega) problems

Explain that Vega can be very small for deep ITM/OTM options, causing large steps or division by near-zero, and that Vega is not always available analytically for some models.

4. Cover step size and convergence issues

Discuss that large steps can overshoot, leading to negative volatility or divergence; small steps may converge slowly. Also mention the need for a good initial guess and bounds.

5. Highlight edge cases and practical solutions

List edge cases: zero or negative implied vol, options with zero Vega, arbitrage violations, and market data issues. Suggest using bracketing methods (e.g., bisection) or hybrid approaches.

Key Points to Mention

  • Newton-Raphson iteration formula and its quadratic convergence when conditions are favorable.
  • Vega as the derivative of the option price with respect to volatility, and its role in the update step.
  • Non-monotonicity of the objective function for deep ITM/OTM options or near expiration.
  • Small Vega leading to numerical instability and large step sizes.
  • Negative volatility or divergence due to overshooting.
  • Practical remedies: good initial guess, bounds, fallback to bisection, or using a hybrid method.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

How would you choose the initial volatility seed before running Newton-Raphson for implied vol, and what heuristics or real data would you lean on to justify that choice?

Technical Trade-offsRoot Cause Analysis
Author's notes

This one I actually felt okay about.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by explaining that the initial volatility seed is critical for Newton-Raphson convergence, especially for deep ITM/OTM options. Then describe a systematic approach: use a model-based heuristic (e.g., Brenner-Subrahmanyam approximation) as a baseline, and refine it with empirical data from similar options (same expiry, near-the-money strikes) or a volatility surface interpolation. Emphasize validation and fallback strategies to handle non-convergence.

Pro tip: Mention that you would implement a hybrid approach: use the Brenner-Subrahmanyam approximation for a quick seed, but if the option is far from the money, switch to a more robust method like the Manaster-Koehler or a bisection step to get a better starting point. This shows you understand both speed and reliability trade-offs.

1. Understand the importance of the seed

Explain that Newton-Raphson can fail or converge slowly if the initial guess is poor, especially for options with low vega or deep ITM/OTM. A good seed reduces iterations and ensures convergence.

2. Use analytical approximations

Describe common heuristics like the Brenner-Subrahmanyam formula (σ ≈ √(2π/T) * C/S) for ATM options, or the Corrado-Miller formula for a more general seed. These are fast and often sufficient.

3. Leverage empirical data and market context

If available, use the implied volatility of nearby options (same expiry, adjacent strikes) or the previous day's implied vol for the same option. For a volatility surface, interpolate from known points.

4. Implement fallback and validation

Set a maximum number of iterations and a tolerance. If Newton-Raphson fails, fall back to a bracketing method like bisection or Brent's method. Validate the seed by checking if the resulting implied vol is within a reasonable range (e.g., 0-200%).

5. Justify with real data and backtesting

Mention that you would backtest the chosen heuristic on historical option data to measure convergence speed and accuracy. Use metrics like average iterations, failure rate, and comparison to a benchmark (e.g., bisection).

Key Points to Mention

  • Brenner-Subrahmanyam approximation for ATM options
  • Corrado-Miller formula for a more robust seed
  • Using the volatility smile/surface for interpolation
  • Fallback to bisection or Brent's method for non-convergence
  • Backtesting heuristics on historical data to evaluate performance
  • Handling edge cases: deep ITM/OTM, near-expiry, low vega

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q3

Write out the closed-form Vega formula for a European option under Black-Scholes, define every variable with units, and explain how Vega's size affects the step size in Newton-Raphson updates.

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

Blanked for a second on the exact form.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by writing the closed-form Vega formula for a European option under Black-Scholes, clearly defining each variable with units. Then explain how Vega's magnitude influences the Newton-Raphson step size when solving for implied volatility, emphasizing that larger Vega leads to smaller adjustments and vice versa.

Pro tip: Mention that near-the-money options have high Vega, so Newton-Raphson converges quickly, while deep in/out-of-the-money options have low Vega, requiring damping or alternative methods to avoid overshooting.

1. Write the Vega formula

Present the closed-form Vega for a European call or put: Vega = S * sqrt(T) * phi(d1), where phi is the standard normal PDF.

2. Define variables and units

Define S (spot price, currency), T (time to maturity, years), phi(d1) (dimensionless), and d1 = (ln(S/K) + (r + sigma^2/2)T) / (sigma * sqrt(T)). Also define K (strike, currency), r (risk-free rate, per year), sigma (volatility, per year).

3. Explain Newton-Raphson for implied volatility

Describe the iterative update: sigma_{n+1} = sigma_n - (BS_price(sigma_n) - market_price) / Vega(sigma_n).

4. Relate Vega to step size

Explain that the step size is inversely proportional to Vega. High Vega (e.g., ATM options) means small adjustments, leading to stable convergence; low Vega (e.g., deep OTM) means large adjustments, risking overshoot and divergence.

5. Discuss practical implications

Mention that for low Vega, one might use damping, bisection, or other robust methods. Also note that Vega is highest for ATM options and decreases as options move away from the money.

Key Points to Mention

  • Vega formula: S * sqrt(T) * phi(d1)
  • Units: S and K in currency, T in years, r and sigma in per year, Vega in currency per unit change in volatility (e.g., per 1% change).
  • Newton-Raphson update uses Vega as the derivative of option price with respect to volatility.
  • Step size = (price error) / Vega, so larger Vega yields smaller steps.
  • High Vega near ATM leads to fast convergence; low Vega deep ITM/OTM can cause instability.
  • Practical fixes: damping, initial guess from ATM, or fallback to bisection.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.