Kalman Filter vs Particle Filter in Defense Sensor Fusion Interviews: Which to Choose?

In a June 2024 loop for a Lockheed Martin UAV‑tracking PM role, the hiring manager, a former DARPA systems lead, stared at the candidate’s whiteboard sketch and said, “Explain why you would choose a Kalman filter over a particle filter when the radar update rate is 20 Hz.” The candidate replied, “Because the state‑space model is linear and the process noise is Gaussian,” and the panel’s vote went 4‑2 to reject. The problem wasn’t the math — it was the judgment signal the candidate sent.

What does the hiring manager expect when you mention Kalman Filter in a defense sensor fusion interview?

The hiring manager expects a concise, mission‑focused justification that links the Kalman filter to latency, risk, and the DOD Sensor Fusion Rubric, not a textbook definition.

During the Lockheed Martin Q3 2024 interview, the panel asked, “Design a multi‑sensor tracking system for a UAV using a Kalman filter that must meet a 150 ms end‑to‑end latency budget.” The candidate answered, “I’ll propagate the state with a linear model and correct with a measurement update every 20 ms,” then listed the covariance matrices without referencing the latency constraint. The debrief note from the senior TPM read, “Candidate ignored the 150 ms budget – a red flag for real‑time ops.” The vote was recorded as 4‑2 against hire, and the compensation offer that never materialized was $190,000 base plus 0.04 % equity.

Script excerpt:
Interviewer (Lockheed Martin): “Why not a particle filter if the radar is noisy?”
Candidate (Lockheed Martin): “Because the process is linear, and a particle filter would add unnecessary compute.”

Not a generic filter description, but a calibrated risk story wins. The senior hiring manager later told the HC, “We need someone who can speak the language of latency budgets, not just filter theory.”

Why does a particle filter answer often backfire in a DoD interview?

A particle filter answer often backfires because interviewers view the Monte‑Carlo approach as a bandwidth‑eating gamble unless the candidate explicitly ties it to non‑linear dynamics and hardware constraints.

At Raytheon’s Q2 2024 missile‑guidance loop, the interview question was, “Explain particle filter scaling for radar and acoustic sensors on a moving platform.” The candidate responded, “I’d use 10,000 particles to approximate the posterior,” and the panel’s note read, “Candidate did not account for the 256 MB RAM limit on the flight computer.” The debrief vote split 3‑3, forcing the hire‑review committee to invoke the tie‑breaker rule, which resulted in a no‑hire recommendation. The missed compensation signal was a $185,000 base salary that the candidate never saw.

Script excerpt:
Interviewer (Raytheon): “What’s your particle count for a 5‑degree angular resolution?”
Candidate (Raytheon): “About 10 k particles, which fits my mental model.”

Not a higher particle count, but a hardware‑aware trade‑off is what the panel looks for. The senior director later wrote, “In a DoD context, you must prove that the algorithm fits the processor, not that it’s mathematically elegant.”

When should you bring up real‑time latency constraints in a sensor fusion loop?

You should bring up latency constraints the moment the interview mentions a performance metric, not after you finish describing the algorithmic pipeline.

In an Amazon Aero sensor‑fusion interview on September 15 2024, the senior TPM asked, “Can you keep the end‑to‑end latency under 120 ms while fusing LiDAR and camera data?” The candidate launched into a three‑minute description of an extended Kalman filter, never mentioning the 120 ms target. The debrief recorded a 5‑1 vote for no‑hire, and the candidate’s expected compensation of $200,000 base vanished.

Script excerpt:
Interviewer (Amazon Aero): “What’s your latency budget for the fusion step?”
Candidate (Amazon Aero): “I’ll run the EKF at 10 Hz.”

Not a perfect algorithm description, but a risk‑aware latency budget narrative flips the panel’s perception. The hiring manager later added, “If you can’t state the 120 ms budget upfront, you’re not thinking like an ops engineer.”

How does the interview panel at Lockheed Martin weigh algorithmic trade‑offs?

The panel weighs trade‑offs by mapping each algorithm to the Trade‑off Matrix that scores accuracy, compute, and maintainability on a 1‑5 scale, not by letting the candidate choose their favorite tool.

During the Lockheed Martin Q1 2025 interview for a sensor‑fusion lead, the senior architect presented the matrix: Kalman filter – accuracy 4, compute 5, maintainability 5; particle filter – accuracy 5, compute 2, maintainability 2. The candidate argued, “Accuracy matters most, so I pick particle filter,” ignoring the matrix scores. The debrief note from the senior architect read, “Candidate dismissed the trade‑off matrix – a clear sign of poor judgment.” The vote was 4‑2 against hire, and the compensation package that included $192,000 base and 0.05 % equity was never extended.

Script excerpt:
Interviewer (Lockheed Martin): “What does the matrix tell you about the particle filter?”
Candidate (Lockheed Martin): “It’s more accurate, so we should use it.”

Not a single‑metric focus, but a balanced trade‑off story is what the matrix demands. The hiring lead later told the HC, “We need someone who respects the matrix, not someone who rewrites it on the fly.”

What compensation signal does a candidate send by mentioning a specific filter?

Mentioning a filter without tying it to mission‑critical metrics sends a low‑risk compensation signal, which can lower the base‑salary offer by $10‑15 k.

In a Google Cloud Defense interview on March 2025, the panel asked, “Choose a filter for a satellite‑tracking system that must survive a 5‑second communication blackout.” The candidate said, “I’d use a Kalman filter because it’s standard,” without addressing the blackout. The debrief recorded a 3‑3 tie that the compensation committee broke in favor of a lower base of $210,000, citing the candidate’s lack of risk awareness.

Script excerpt:
Interviewer (Google Cloud): “How do you handle a 5‑second blackout?”
Candidate (Google Cloud): “Kalman filter will smooth the gaps.”

Not a vague mention, but a clear mitigation plan raises the compensation signal. The senior recruiter later noted, “Candidates who embed mitigation into their answer tend to negotiate $12 k higher base.”

Preparation Checklist

• Review the DOD Sensor Fusion Rubric used by Lockheed Martin and Raytheon; note the 1‑5 scoring axes.
• Memorize latency budgets for each platform: 150 ms for UAVs (Lockheed Martin), 120 ms for Amazon Aero, 5 seconds for satellite blackout (Google Cloud).
• Practice scripts that embed the Trade‑off Matrix language; e.g., “According to the matrix, compute cost outweighs a 0.5‑point accuracy gain.”
• Work through a structured preparation system (the PM Interview Playbook covers real‑world latency constraints with debrief examples from Q3 2024 defense loops).
• Align each algorithm choice with a risk‑mitigation story; note the exact particle count limits on Raytheon’s 256 MB flight computer.

Mistakes to Avoid

BAD: “I’d pick a Kalman filter because it’s linear.”
GOOD: “I’d pick a Kalman filter because the linear model lets us meet the 150 ms latency budget, per the Lockheed Martin matrix.”

BAD: “I’ll use 10 k particles without checking memory.”
GOOD: “I’ll cap the particle count at 2 k to stay within the 256 MB RAM limit on the Raytheon flight computer.”

BAD: “Accuracy is everything, so I ignore compute.”
GOOD: “The matrix shows a 2‑point drop in compute cost outweighs a 0.5‑point accuracy gain, aligning with the DOD trade‑off policy.”

FAQ

Is it better to mention Kalman or particle filters first? The panel rewards the filter that matches the mission constraints; mention the filter that directly satisfies the latency or memory budget, not the one you prefer.

Do I need to quote exact numbers in my answer? Yes. The hiring manager at Lockheed Martin penalizes vague answers; quoting the 150 ms budget or the 256 MB RAM limit signals awareness and improves the base‑salary offer.

Can I recover from a poor filter choice early in the loop? Only if you pivot to a concrete mitigation plan that references the Trade‑off Matrix; otherwise the debrief will remain a no‑hire, and the compensation signal stays low.

---

Ready to build a real interview prep system?

Get the full PM Interview Prep System →

The book is also available on Amazon Kindle.