← Upstart Interview Insights

Upstart·Data Scientist·Technical Phone Screen·Intermediate

Intermediate
Jul 2026

Summary

Upstart data scientist interview that leaned pretty hard into probability theory. The core question was a physics-flavored stats problem and it required you to actually derive something, not just name-drop a distribution.

Questions Asked (1)

Q1

You have 100 independent particles, each with the same known half-life. Derive the probability that exactly k particles (or at least one) are still undecayed after some time t.

Algorithms & Data StructuresProduct Analytics & Metrics
Author's notes

The setup sounds like a physics problem but it's really just asking you to chain together the exponential survival function and binomial probability.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by recognizing that each particle's decay is an independent Bernoulli trial with probability p = e^{-λt} of surviving. Then model the number of survivors as a Binomial(n=100, p) random variable and derive the probability mass function for exactly k and the complement for at least one. Finally, connect the result to the exponential decay law and discuss implications for large n.

Pro tip: Explicitly state the assumptions (independence, identical half-life) and note that the binomial model is exact for any n, but for large n you can approximate with Poisson or normal—this shows you understand both theory and practical scaling.

1. Define the per-particle survival probability

Given half-life T, the decay constant λ = ln(2)/T. The probability a single particle survives time t is p = e^{-λt} = 2^{-t/T}.

2. Model the number of survivors

Since particles are independent and identical, the number still undecayed after time t follows a Binomial distribution with parameters n=100 and p as above.

3. Derive probability for exactly k survivors

Use the binomial PMF: P(X=k) = C(100, k) * p^k * (1-p)^{100-k}, where C(100,k) is the binomial coefficient.

4. Derive probability for at least one survivor

Use the complement: P(X ≥ 1) = 1 - P(X=0) = 1 - (1-p)^{100}.

5. Discuss extensions and approximations

Mention that for large n and small p, the Poisson approximation with λ = np can be used, and for large n, the normal approximation applies. Also note the expected number of survivors is np.

Key Points to Mention

  • Independence and identical distribution of particles
  • Relationship between half-life and decay constant: λ = ln(2)/T
  • Binomial distribution for the count of survivors
  • Binomial coefficient and PMF formula
  • Complement rule for 'at least one' probability
  • Poisson or normal approximation for large n (optional but shows depth)

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