Microsoft Security & Identity intern: resume questions

Tue Nov 10, 2026. Two 45-minute rounds on Teams (1:00 and 2:00 PM ET), each with resume and behavioral talk, then a LeetCode medium in HackerRank. This site is the resume half. Every question has an answer you can open. Anything marked check is my best guess from your notes; fix it if it is wrong.

What the resume part looks like

Rules for every answer

The 60-second intro

"I'm Rishabh, a CS student at Penn State, graduating 2028. Last summer I interned at Cisco on AI infrastructure. I extended the production simulator for their cloud-managed data center fabric in Go and Python: device telemetry for switches, servers and BMCs, live port breakout on a virtual Linux network, and a proposal I ran with seven engineers for simulating customer servers end to end, including a software TPM for device identity. Before that I built internal AI tools and an MCP server at a bank, so I spent a summer inside compliance reviews. On my own I built Atrax, a platform where coding agents deploy apps with their own databases and URLs, and that is where I kept running into the problem I want to work on: identity and permissions for things that are not people. That is why Security and Identity."

Keep the IIT Madras degree for the education question only. Do not open with hackathons or TrendFlip.

Drill sprints

Same as the coding sprints: one question per column, out loud, 60-second timer, then the next. Three tonight, three Monday, Sprint 0 as the Tuesday warm-up. Open the answer only after you have said yours.

SprintIntro / behavioralCiscoMashreq or AtraxProjects / open sourceSecurity angle
0 (Tue)60-second introServer proposal in 60 sAtrax in 30 sKueue bug in 30 sWhy Security & Identity
1Tell me about yourselfHow did you measure 0% dropsWhere do Atrax DB credentials liveNotionBench gradingWhat is the TPM for
2A time you disagreedWhy not mount the Docker socketHow did the bank tools stay groundedKueue: two identifiers for one nodeDetecting garbage telemetry
3A failure and what changed60% of what, and what did a test catchMigration fails halfwayPDFSpacer production incidentWho is the identity in Atrax
4Influence without authorityRunning the review with 7 peopleWho could call the MCP serverGo vs Python vs TypeScriptFidelity limit of the nested design
5Why MicrosoftWhat is a BMC and why mock itHow are Atrax apps isolatedWhat surprised you in NotionBenchWhat would you not let an agent touch
6Where do you want to growBreakout race conditionSensitive data at the bankShare-Pen: why did you sellAgent identity in two sentences

Only you can answer these

Each is marked check where it appears. Five minutes each, written down, before Tuesday.

  1. Breakout: watcher mechanism, interface types, exact tool and duration behind "0% dropped."
  2. 60%: exact weeks planned vs actual, and one real thing an agent got wrong that a test caught.
  3. Telemetry: the hardest fixture and why.
  4. Server proposal: the one contested open question and how it was resolved.
  5. Mashreq: where the models ran, what left the bank's boundary, how MCP callers authenticated, how the portal did login and roles.
  6. Atrax: per-app credentials or shared, migration bookkeeping, how tenants are isolated today, who holds the deploy token.
  7. One real failure with the change you made after it.
  8. One PDFSpacer production incident.
  9. One NotionBench finding that surprised you.
  10. Whether you ever saw real Hyperfabric hardware or only the simulator. Say it plainly either way.