4. Round 4, Bar Raiser, HM

What can happen after a third live. Job 10454435 still none. AUTA India 2025–2026 first-hand posts are the source. Labels: IE-asked.

A third live is not automatically the end

Accounts that say what came next, after two lives plus another hour, still show a later Bar Raiser or HM:

I do not have AUTA-fresher evidence of a fifth live as a normal step. Absence is not a promise Amazon cannot add one.

If the interviewer is a Bar Raiser

You may hear that they interview across orgs, or that they are not on the hiring team. You may hear nothing. Do not announce “so this is the Bar Raiser.” Do the hour.

What candidates actually got: long LP, metrics, “what would you do differently,” sometimes still one coding question. Nisarg AUTA BR went through many LP STARs. GFG 2025 BR (not necessarily UTA) asked current role, last deep-dive, conflict, last negative feedback, learned something not required — those prompts are in bible §5.D and answered in chapter 08.

Keep one Java DSA warm: tree path, Distance K, Nice Subarrays / prefix, a heap, or Maximum Sum Linked List (switch at common nodes). Bar Raiser coding showed up often enough that dropping Live Code mode is a mistake. Keep genuine STAR warm: chapter 12. If they visit a deployed app, only defend the resume.

If they ask about missing a deadline: I do not have a missed ERC or a missed 99.9 percent on the resume, and I will not invent one. What I will speak is how an immovable date was met — 60 FPS and a live costmap sequenced first — and how I would communicate if a ship-blocker were late: early, specific, cut extra hardware, do not cut isolation, do not scale a 73 percent safety model.

If the interviewer is an HM / SDM

LC 7563011 is the AUTA template: in-depth research, POCs, multiple solutions — IQVIA Firecrawl vs one scraper, vector vs hybrid+GraphDB, evals as the stop condition. They will 10x the users and ask why this database. That is chapter 09, not Kafka HLD.

Bhavya’s Spring login Controller / Facade / Service / Repository was an HM design, not the unverified Login Tracker, not R2 LRU.

GenAI on an HM hour: LC 7724048 and LC 8014509. Verify, do not worship the model. Same hour may still paste Next Greater or LCA.

Decision tree (only as far as evidence goes)

Your R3 hour
   ├─ mostly LP / HM / project  → often closer to a final debrief
   ├─ DSA / Fluency-as-DSA / LLD → a Bar Raiser after is common in AUTA posts
   └─ mixed                     → stay ready for either an offer wait or another invite

Do not treat a strong R3 as the finish line. Do not treat a coding R3 as a failure to reach BR. Mixed feedback is why extra hours exist.

Questions to ask them (end of R3 or R4)

Do not ask compensation in this loop stage. Do not ask “what does Amazon do.”