Tell me about a time you owned a system end-to-end.
“Invented a DC-DC converter → 25 % space efficiency ↑, 50 % cost ↓, saved $100 k NRE.”
STAR
S Each onboard computer carried four DC-DC bricks, driving $4k cost and 200 grams mass per unit
T Design a single DC-DC Power Converter (UDPC) that outputs 5 V, 24 V rails at ≥90 % efficiency.
A Selected flyback topology, sized custom magnetics using RidleyWorks, ran LT-spice Monte-Carlo (500 corners), laid out 4-layer PCB got it manufactured then tested three prototypes in-house
R UDPC hit 92 % efficiency, 50 % cost reduction, 75 % area shrink; default on all on board computers.
LPs
Invent & Simplify | Dive Deep | Deliver Results | Frugality | Ownership
Tell me about a time you had incomplete information but had to act.
“Single star-ground connection → comms restored instantly, zero hardware changes, saved 3 month delay”
S – While testing the PCDU and OBC together, communication failed initially even though everything was set up how the test plan had it.
T – I needed to restore communication so testing could proceed.
A – I reviewed the PCDU specification and noticed a star-grounding requirement. I connected the chassis of the power supply, PCDU, and OBC according to the grounding diagram.
R – Communication immediately began working, testing resumed, I updated the test procedure, and the issue was resolved without hardware changes, saved 3 months by not changing out the units.
LPs
Dive Deep | Are Right, A Lot | Earn Trust | Knowledge Sharing
Tell me about a time you had to dive deep to solve a difficult problem.
“Invented a DC-DC converter → 25 % space efficiency ↑, 50 % cost ↓, saved $100 k NRE.”
STAR
S Each onboard computer carried four DC-DC bricks, driving $4k cost and 200 grams mass per unit
T Design a single DC-DC Power Converter (UDPC) that outputs 5 V, 24 V rails at ≥90 % efficiency.
A Selected flyback topology, sized custom magnetics using RidleyWorks, ran LT-spice Monte-Carlo (500 corners), laid out 4-layer PCB got it manufactured then tested three prototypes in-house
R UDPC hit 92 % efficiency, 50 % cost reduction, 75 % area shrink; default on all on board computers.
LPs
Invent & Simplify | Dive Deep | Deliver Results | Frugality | Ownership
Tell me about a time you improved the customer experience.
“Data-driven feeder analysis → 15 % SAIDI & SAIFI ↓, Python dashboard now used by 3 operating regions.”
STAR
S SAIDI 118 min, SAIFI 1.4—regulatory warning; repeat failures on specific feeders.
T Cut indices 10 % within 12 months (regulatory commitment).
A Pulled 5-year outage db, k-means clustering by cause code, weather, device age, built Python dashboard that auto-flags >2× median feeders; field-validated 18 sites, prioritized recloser upgrades, vegetation trims.
R SAIDI 118 → 100 min, SAIFI 1.4 → 1.18, 15 % improvement; dashboard live every Monday, 3 regions adopted, prevented 900 k customer-minutes last year.
LPs
Customer Obsession | Dive Deep | Deliver Results | Invent & Simplify
Tell me about a time you refused to accept “good enough.”
“Raised begin-charge set-point 10 °C → 15 °C → 30 % cold-start risk ↓, zero mass impact, adopted as standard for 3 other programs.”
STAR
S 10 °C begin-charge met spec, but left 1 °C margin after degradation—“good enough”.
T Increase robustness without new heaters or schedule slip.
A Re-ran python energy balance model with battery heaters running until the battery reached 15 °C, also made sure to include the extra time in the TVAC testing
R Cold-start risk ↓ 30 %, fleet-wide update uploaded; now baseline on three follow-on satellite constellations.
LPs
Insist on Highest Standards | Ownership | Are Right, A Lot | Customer Obsession
Tell me about a time you had to learn something completely new.
“Intern, zero plasma background → taught myself tool settings in 1 week → cut defect rate, pushed yield from 60 % to 90 %.”
STAR
S On day-one the fab’s DRIE tool was scraping 40 % of wafers; yield stuck at 60 % and engineers were booked solid.
T My mentor said, “Run a design-of-experiments on the tool—don’t break anything.”
A I had never seen a generator panel, 480 V breaker, or VFD, so the facilities lead walked me through the full path: emergency generator → ATS → 480 V switchgear → breaker → VFD → pump motor. I built a 16-run DOE matrix—pump Hz 35-60, bias V 150-180, pressure 4-8 mTorr—and logged defect count for every wafer. The data showed 42 Hz + 165 V gave the smoothest etch; I ran a 5-wafer confirmation lot.
R Defects fell by half, yield jumped from 60 % to 92 %, and the DOE recipe is still the standard—all from an intern who learned the generator-to-tool power chain and DOE in under a week.
Tell me about a time you had to make a judgment call with tradeoffs
“Invented a DC-DC converter → 25 % space efficiency ↑, 50 % cost ↓, saved $100 k NRE.”
STAR
S Each onboard computer carried four DC-DC bricks, driving $4k cost and 200 grams mass per unit
T Design a single DC-DC Power Converter (UDPC) that outputs 5 V, 24 V rails at ≥90 % efficiency.
A Selected flyback topology, sized custom magnetics using RidleyWorks, ran LT-spice Monte-Carlo (500 corners), laid out 4-layer PCB got it manufactured then tested three prototypes in-house
R UDPC hit 92 % efficiency, 50 % cost reduction, 75 % area shrink; default on all on board computers.
LPs
Invent & Simplify | Dive Deep | Deliver Results | Frugality | Ownership
Tell me about a time you thought beyond your immediate task.
“Wrote 20-line Python script → identified 8 reaction wheels in 5 min, cut integration time 25 %, now in official Docker image.”
STAR
S Flat-SAT integration blocked because reaction-wheel CAN addresses unknown; manual brute-force >2 days.
T Keep schedule (launch campaign fixed).
A Scripted incremental CAN sweep 0x01-0xFF, spun each wheel 5s at 5 % max torque, once the wheel was identified, the address was reconfigured to the correct address
R 8 wheels identified in 5 min, integration finished 1 day early;
Tell me about a time you made a mistake.
“DC-DC card passed no-load, sagged under full load; swapped to lower-ESR caps, saved four-month respin, checklist now guards against it.”
STAR
S My UDPC fly-back converter passed bench tests at zero load, so I shipped the first article to system integration.
T During full-load burn-in the 5 V rail sagged to 4.5 V, risking digital resets—and a four-month board respin.
A I admitted I’d only validated light load, then isolated blocks with an electronic load: the feedback compensation was too conservative for the ESR zero I’d modeled. I replaced the 80 mΩ output capacitors with 25 mΩ low-ESR parts, doubling the zero frequency and restoring tight regulation; re-tested 0–6 A sweep—5 V stayed within 1 %. I updated the power checklist to “step-load full current before release.”
R Card worked perfectly, no respin, and the checklist now blocks the same oversight on every new design.
Leadership Principles
Ownership | Learn and Be Curious | Insist on Highest Standards | Deliver Results
Tell me about a time you disagreed with a technical decision.
“Disagreed with 10 °C battery start temp—proved 15 °C safer, got team flip, cut cold-start risk 30 %.”
STAR
S The battery team wanted to keep the 10 °C begin-charge threshold because “it meets spec and we’re out of schedule margin.”
T As the junior power engineer I had to prove a higher limit was worth the schedule hit or accept the risk.
A I pulled 5-year orbital thermal data, showed 15 % of cycles could dip below 10 °C after degradation; built a quick Monte-Carlo and ran 8 extra TVAC cycles at 15 °C (after-hours). Heater duty only rose 4 %, still inside power budget. I presented the 30 % risk-reduction figure at the weekly CCB.
R The team reversed the decision, raised the set-point to 15 °C, and the cold-start risk was cut 30 %—no schedule slip, no new hardware.
Leadership Principles
Have Backbone; Disagree & Commit | Dive Deep | Customer Obsession | Deliver Results
Tell me about a time you reduced cost or saved money.
“Invented a DC-DC converter → 25 % space efficiency ↑, 50 % cost ↓, saved $100 k NRE.”
STAR
S Each onboard computer carried four DC-DC bricks, driving $4k cost and 200 grams mass per unit
T Design a single DC-DC Power Converter (UDPC) that outputs 5 V, 24 V rails at ≥90 % efficiency.
A Selected flyback topology, sized custom magnetics using RidleyWorks, ran LT-spice Monte-Carlo (500 corners), laid out 4-layer PCB got it manufactured then tested three prototypes in-house
R UDPC hit 92 % efficiency, 50 % cost reduction, 75 % area shrink; default on all on board computers.
LPs
Invent & Simplify | Dive Deep | Deliver Results | Frugality | Ownership
Tell me about a time you simplified a process.
“Invented a DC-DC converter → 25 % space efficiency ↑, 50 % cost ↓, saved $100 k NRE.”
STAR
S Each onboard computer carried four DC-DC bricks, driving $4k cost and 200 grams mass per unit
T Design a single DC-DC Power Converter (UDPC) that outputs 5 V, 24 V rails at ≥90 % efficiency.
A Selected flyback topology, sized custom magnetics using RidleyWorks, ran LT-spice Monte-Carlo (500 corners), laid out 4-layer PCB got it manufactured then tested three prototypes in-house
R UDPC hit 92 % efficiency, 50 % cost reduction, 75 % area shrink; default on all on board computers.
LPs
Invent & Simplify | Dive Deep | Deliver Results | Frugality | Ownership
Tell me about a time you stepped outside your formal role to help the team succeed.
“Wrote 20-line Python script → identified 8 reaction wheels in 5 min, cut integration time 25 %, now in official Docker image.”
STAR
S Flat-SAT integration blocked because reaction-wheel CAN addresses unknown; manual brute-force >2 days.
T Keep schedule (launch campaign fixed).
A Scripted incremental CAN sweep 0x01-0xFF, spun each wheel 5s at 5 % max torque, once the wheel was identified, the address was reconfigured to the correct address
R 8 wheels identified in 5 min, integration finished 1 day early;
Tell me about a time you used data to identify root cause.
“Data-driven feeder analysis → 15 % SAIDI & SAIFI ↓, Python dashboard now used by 3 operating regions.”
STAR
S SAIDI 118 min, SAIFI 1.4—regulatory warning; repeat failures on specigiv feeders.
T Cut indices 10 % within 12 months (regulatory commitment).
A Pulled 5-year outage db, k-means clustering by cause code, weather, device age, built Python dashboard that auto-flags >2× median feeders; field-validated 18 sites, prioritized recloser upgrades, vegetation trims.
R SAIDI 118 → 100 min, SAIFI 1.4 → 1.18, 15 % improvement; dashboard live every Monday, 3 regions adopted, prevented 900 k customer-minutes last year.
LPs
Customer Obsession | Dive Deep | Deliver Results | Invent & Simplify
Tell me about a time you took a calculated risk.
Solar-string short menaced mission—MATLAB energy-balance across BOL, EOL, orbit-raising, station-keeping, safe modes & de-orbit proved 6 % margin, waived $2 M rework, users kept imaging on schedule.”
S After delivery, one solar cell shorted on a commercial imaging satellite; customer faced $2 M removal/rework and 14-month delay—risking their launch slot and insurance.
T As power intern, prove the satellite is still safe for operators and end-users or the mission stops.
A I wrote a MATLAB energy-balance script that stepped through every mission phase: Beginning-of-Life (BOL) sun-pointing, End-of-Life (EOL) with 3 % cell degradation, orbit-raising with 45 ° off-pointing, station-keeping with 10 ° seasonal swings, safe-mode power-saving, and final de-orbit battery-only ops. In each mode I subtracted the lost substring power, added battery SOC change, and iterated over a 5-year orbital lifetime. Output showed only 2.1 % energy loss with 6 % margin remaining. I packaged the plot and a one-page brief stating “no impact to imaging duty cycle or ground contacts” and presented it to the customer’s ops team.
R Customer accepted the waiver, skipped the rework, and the satellite launched on time. End-users now receive uninterrupted imagery, and the MATLAB energy-balance script is the fleet standard for any future string-loss decision.
Leadership Principles
Customer Obsession | Are Right, A Lot | Deliver Results | Earn Trust
Tell me about a time you protected the end user.
“Solar-string short menaced mission—MATLAB energy-balance across BOL, EOL, orbit-raising, station-keeping, safe modes & de-orbit proved 6 % margin, waived $2 M rework, users kept imaging on schedule.”
S After delivery, one solar cell shorted on a commercial imaging satellite; customer faced $2 M removal/rework and 14-month delay—risking their launch slot and insurance.
T As power intern, prove the satellite is still safe for operators and end-users or the mission stops.
A I wrote a MATLAB energy-balance script that stepped through every mission phase: Beginning-of-Life (BOL) sun-pointing, End-of-Life (EOL) with 3 % cell degradation, orbit-raising with 45 ° off-pointing, station-keeping with 10 ° seasonal swings, safe-mode power-saving, and final de-orbit battery-only ops. In each mode I subtracted the lost substring power, added battery SOC change, and iterated over a 5-year orbital lifetime. Output showed only 2.1 % energy loss with 6 % margin remaining. I packaged the plot and a one-page brief stating “no impact to imaging duty cycle or ground contacts” and presented it to the customer’s ops team.
R Customer accepted the waiver, skipped the rework, and the satellite launched on time. End-users now receive uninterrupted imagery, and the MATLAB energy-balance script is the fleet standard for any future string-loss decision.
Leadership Principles
Customer Obsession | Are Right, A Lot | Deliver Results | Earn Trust
Tell me about a time you worked outside your core expertise.
“Wrote 20-line Python script → identified 8 reaction wheels in 5 min, cut integration time 25 %, now in official Docker image.”
STAR
S Flat-SAT integration blocked because reaction-wheel CAN addresses unknown; manual brute-force >2 days.
T Keep schedule (launch campaign fixed).
A Scripted incremental CAN sweep 0x01-0xFF, spun each wheel 5s at 5 % max torque, once the wheel was identified, the address was reconfigured to the correct address
R 8 wheels identified in 5 min, integration finished 1 day early;
Tell me about a time you raised the quality bar.
“Dropped DRIE electrode bias 180 V → 165 V + pump VFD tune → 7 % faster etch, uniformity ±3 % → ±2 %, yield 91 % → 98 %, +1 extra wafer/shift, recipe now fleet standard.”
STAR
S ARL’s Deep Reactive-Ion Etcher showed 91 % die yield and plateaued etch rate on 6 µm-thick SOI MEMS runs; bias held fixed 180 V, pump 60 Hz constant.
T Boost yield to >95 % and increase daily throughput without hardware spend.
A Designed DOE matrix (bias, pressure, VFD speed); lowered electrode bias 180 V → 165 V while ion density unchanged (Langmuir probe), tightened uniformity ±3 % → ±2 %; sped VFD 60 Hz → 42 Hz to match new pressure set-point, cut micro-loading defects 40 %.
R Etch rate ↑ 7 %, yield 91 % → 98 %, +1 wafer/shift (+8 % weekly output); recipe uploaded to MES, now baseline across three fabs, zero capex.
Leadership Principles
Insist on Highest Standards | Deliver Results | Dive Deep | Invent & Simplify
Tell me about a time you built something reusable.
Invented a DC-DC converter → 25 % space efficiency ↑, 50 % cost ↓, saved $100 k NRE.”
STAR
S Each onboard computer carried four DC-DC bricks, driving $4k cost and 200 grams mass per unit
T Design a single DC-DC Power Converter (UDPC) that outputs 5 V, 24 V rails at ≥90 % efficiency.
A Selected flyback topology, sized custom magnetics using RidleyWorks, ran LT-spice Monte-Carlo (500 corners), laid out 4-layer PCB got it manufactured then tested three prototypes in-house
R UDPC hit 92 % efficiency, 50 % cost reduction, 75 % area shrink; default on all on board computers.
LPs
Invent & Simplify | Dive Deep | Deliver Results | Frugality | Ownership
Amazon Leadership Principles Highlighted:
Invent and Simplify
Deliver Results
Dive Deep
Data Center Electrical One-Line Diagrams
Utility feed (e.g., 34.5 kV / 24 kV)
Medium-Voltage switchgear
Step-down transformers (34.5 kV → 480 V)
Low-voltage switchgear
UPS topology (centralized vs distributed)
Generator integration
ATS vs STS
Power Distribution Units (PDUs)
Redundant paths (A/B feeds)
Amazon Leadership Principles
Customer Obsession: Start with the customer and work backward.
Ownership: Think long-term and don’t sacrifice long-term value for short-term results.
Invent and Simplify: Innovate and simplify, and never say “no” to an idea because of a belief that it can’t be done.
Are Right, A Lot: Leaders are right a lot, have strong judgment, and seek diverse perspectives.
Learn and Be Curious: Always be learning and seeking to improve themselves.
Hire and Develop the Best: Raise the performance bar with every hire and promotion.
Insist on the Highest Standards: Drive continuous improvement and demand excellence.
Think Big: Think big and have a bias for action.
Bias for Action: Speed matters in business; many decisions and actions are reversible.
Frugality: Accomplish more with less; constraints breed resourcefulness.
Earn Trust: Listen attentively, speak candidly, and treat others respectfully.
Dive Deep: Operate at all levels, stay connected to the details, and audit frequently.
Have Backbone; Disagree and Commit: Respectfully challenge decisions when you disagree, then commit wholly once a decision is made.
Deliver Results: Focus on key inputs and deliver them with the right quality and timeliness.
Strive to be Earth’s Best Employer: Create a safer, more productive, more diverse, and more just work environment.
Success and Scale Bring Broad Responsibility: Think big and act on behalf of the entire company.
-
CBP Interview Perep16
-
More CBP stuff19
-
SENSRX23
-
Filters2
-
Flyback31
-
linear-reg & reference questions8
-
switches4
-
Amplifiers3
-
Surge Protection3
-
WCCA2
-
L3 Harris actual interview29
-
SILLT37
-
Engineering Tech65
-
TUNNEL9
-
AST SpaceMobile22
-
KCW15
-
KFORM14
-
GD 2nd interview31
-
Amazon Comissioning23
-
Amazon Commissioning 221
-
kimi k amazon7
