This is a rehearsal notebook for Adarsh Vishwakarma. Read it as yourself. R1 was 13 Aug 2026. R2 was 18 Aug 2026. R3 is done. This chapter is a 60-minute spine from other leftover hours — not a forecast that yours will match the clock. Live Code is Java. Job 10454435 still has no public write-up that names a live-round question. Unnamed stays unnamed.
The calendar is usually one hour. Other leftover write-ups used that hour as intro plus a few deep LPs, as thirty/thirty LP and DSA, as DSA first then a couple of LPs, or as work talk then leftover code because time remained. Five of eleven locked hours ended in about 22–40 minutes. You still prepare the full hour. n=11 is not a population and not this Job ID.
What they might ask
Tell me about yourself in the first two minutes.
Three or four Leadership Principle stems with follow-ups on the same story.
A split: ~30 LP / ~30 DSA, or DSA ~30 then a couple LP.
Two interviewers in one hour.
“Do you have questions?” in the last few minutes.
Hour dump, 11 must-opens. IE-asked as shape. Zero public hours for this Job ID. Nisarg claimed all 16 Leadership Principles at ~15–20 minutes each and a 15:00–16:00 Pacific hour. Those two statements cannot both be literal. Leave the clash as a clash. Few-deep is the common write-up (9/11 in that lock).
What you say
I’m ready to start whenever you are.
They might ask next
If they ask
You answer
Two interviewers split the hour
Answer the person who asked. Recap the intro only if the second person asks.
They skip intro and paste a problem
Skip to Java. Restate constraints.
They never fill sixty minutes
Ask team-week questions, thank them, stop. Early end showed on 1qj304z, 7623949, 6425074, IE.in, and 7563011.
Time / what often happens / what you do
Time
What often happens
What you do
0:00–0:02
Names. They may ask tell-me-about-yourself.
Speak the intro below. Stop. Six projects wait until they pick a door.
LP-heavy (other loops)
3–4 stems, follow-ups on the same story. LC 6806195 three named + follow-ups on all; IE.in 2024-grad five named, ~37 min; LC 8362604 four topic families, first 15–20 min intense; LC 7623949 intro + Why Amazon + three LP names, ~22 min.
One project per stem. Answer why on that story. Default pair if they name nothing: Dive Deep then Deliver.
Mixed
~30 LP / ~30 DSA (6806195) or DSA ~30 min then couple LP (LC 7406809).
Finish the open stem, then Java if the editor opens. Couple LP after code still gets a real mechanism.
Work then leftover
~40 min stack/architecture, then leftover DSA (Shiwangi R4, 17:00–18:00).
After work talk, restate constraints and write Java. Unnamed sliding window stays unnamed.
Answer the person who asked. Do not restart STAR for the second person unless they ask a new stem.
Last 3–5 min
“Do you have questions?”
Team’s week for a new SDE I. Compensation stays off this hour.
Keep straight
This clock is a reading aid. You do not narrate minutes out loud.
Padding to sixty minutes with a sixth project is not the leftover default. StratifyLabs, GiftedBooks, and Argus wait until they pick a project or an LP that maps there.
An early end is a write-up fact on five of eleven locked hours. It is not a pass/fail script.
Tighten
Follow the row they actually opened. If they skip a block, skip it. Full spoken STARs are chapter 3.
Ninety-second intro
Tell me about yourself. Current work, the three internships as a timeline, then a door they can pick. Current role is IQVIA. If they cut you at forty-five seconds, name + CUSAT + IQVIA already stand.
What they might ask
Tell me about yourself.
Walk me through your resume.
Introduction, then Why Amazon (LC 7623949 leftover BR opener shape).
GFG 2025 Bar Raiser opened with current working role. LC 7623949 leftover BR opened with Introduction then Why Amazon. LC 6881058 had a current-role discussion; exact sentence unnamed. IE-asked as an opener shape. Resume-derived for the words below. Spoken paragraph matches the home page.
What you say
I am Adarsh Vishwakarma. I finish a B.Tech in CSE at CUSAT in May 2026 (CGPA 8.42). I am currently a Software Developer Intern at IQVIA in Kochi, April 2026 to present, on two FastAPI systems: LangGraph Deep Research across 200+ websites, and Hybrid RAG on 200+ page BRDs with Azure AI Search, GraphDB, and LangSmith. Before that, at Ylogx from November 2024 to October 2025 I shipped production BI: reports, SQL RAG, three-tier RLS, Redis, and AWS. Earlier, from February to June 2024, I was core software on Team Horizon’s ROS2 rover, 17th of 80+ at ERC 2024. I can go deep on IQVIA retrieval, Ylogx isolation, or Horizon’s live costmap. Which do you want?
They might ask next
If they ask
You answer
The second IQVIA system in more detail
Hybrid RAG on Azure AI Search — hybrid and semantic retrieval plus a GraphDB — for two-hundred-page BRDs, with LangSmith tracing and generated tests. Still IQVIA. Still not Ylogx.
They cut you at forty-five seconds
The first three sentences already stand. Horizon, Argus, GiftedBooks, StratifyLabs, and eight hackathons wait.
They already heard IQVIA in a previous live
Lead one failure mode: two high-ranked pages disagree, or hybrid versus vector-only. Then stop. Architecture retell waits.
Which intern ran longest in production
Ylogx, November 2024 to October 2025. That is a date.
Keep straight
One intro, three internships as a timeline. StratifyLabs, GiftedBooks, and Argus are projects; they wait until they pick a project.
Ylogx 40 / 99.9 / +65 / −35 / 30 KPI / sub-210 stay off the intro. Those families belong on Ylogx LP stories.
17th/80+ is in the intro as Horizon’s one line. A full rover STAR waits until they pick that door.
Internships used Python, TypeScript, and ROS2. Live Code is Java. Those internships are not JVM/Spring services.
No Amazon team name on the resume. None in the intro.
Tighten
End on “Which do you want?” and wait. Current-role-only wording is the next card if they skip the timeline.
Current role
Current working role and responsibility is a different stem from tell-me-about-yourself. It is still not an LP. Two systems. Resume numbers only. Stop. IQVIA is April 2026 to present, Kochi.
What they might ask
Current working role and responsibility (GFG 2025 BR, printed).
What do you do right now?
Walk through current project architecture (LC 6866029 leftover-shaped architecture stem).
IE-asked GFG 2025 BR. OA-as-R1 on that loop — the BR after two technicals is leftover-shaped. Same lock as the intro: IQVIA April 2026–present, Kochi. Resume-derived.
What you say
I am a software developer intern at IQVIA in Kochi, April 2026 to present. I work on two systems. The first is LangGraph Deep Research across more than two hundred websites. The second is Hybrid RAG: Azure AI Search hybrid and semantic retrieval, a GraphDB, LangGraph choosing retrieval, LangSmith for evals and generated test cases on more than two-hundred-page BRDs.
They might ask next
If they ask
You answer
What did you do before
One sentence Ylogx (November 2024–October 2025: FastAPI, NestJS, Postgres, SQL RAG, three-tier RLS, Redis, ECS) or one sentence Horizon (February–June 2024: ERC 17th of 80+). Then wait.
Whether this intern is Java
Production here is Python FastAPI. I write Java in Live Code.
Architecture as a later stem
This same walkthrough, not a new LP trophy.
They already heard IQVIA
Switch to Ylogx FastAPI + NestJS + Postgres RLS + Redis + ECS/ALB/CloudFront as the systems you owned. Still not a fourth IQVIA primary.
Keep straight
GraphDB stores requirement dependencies. LangGraph is workflow state.
Spring is not on the IQVIA bullets. Production there is Python FastAPI.
−35% bot DB and sub-210 ms are Ylogx. IQVIA has no latency percent on the resume.
Current role is not Ownership / Invent / Learn spoken as three IQVIA trophies in the intro.
Tighten
Stop after two systems. If they want a principle, chapter 3. If they want mechanisms, chapter 5.
Invite said LP-only
The recruiter or the mail may say Leadership Principles only, or the invite may lack a Live Code link. The interviewer can still paste a tree or an array. You do not decide the round from the calendar. Two AUTA third lives labeled Bar Raiser were selected after that miss: 7406809 and 6806195.
What they might ask
A tree path, print A/B/C, extra space banned (LC 7406809). HR had said only three LP questions.
Count Number of Nice Subarrays after 30 minutes of LP (LC 6806195). Mail had no Live Code link.
LP only, matching the invite (Reddit 1qj304z; LC 7623949 format shared beforehand).
AUTA n=2 where framing implied no DSA and the room had DSA — both third lives labeled Bar Raiser, both selected. Matching LP: Reddit 1qj304z and LC 7623949. Matching is not permission to unprep Java. IE-asked as a class, not a rate, not this Job ID. GFG 2025 “unsure because of LP” was their second live (already-done analogue).
What you say
I’ll restate the problem in one sentence, ask about nulls, duplicates, and n, name a brute that is too slow, then write Java.
They might ask next
If they ask
You answer
Extra space is banned on a tree
LCA plus DFS, not a parent HashMap.
They switch from LP to code
Leave the STAR. Editor first.
They never open an editor
Stay on stories.
Couple LP after code
Ylogx Redis + RLS or Horizon 17th — a real mechanism, not a slogan.
Keep straight
The recruiter promised no coding is not a sentence you say in the room.
LP still gets mechanisms even when a write-up called it formality.
Live Code is Java. Internship production is not the editor language.
Tighten
Chapter 1 has the framing table. This card is the switch: story until they paste, Java when they paste.
All 16 versus three or four deep
Some leftover hours go wide. Most go deep on a few stems. One AUTA LinkedIn post claimed all sixteen Leadership Principles, each explored 15–20 minutes, in a one-hour calendar slot. You cannot literally do both. You go deep on the prompt they asked. Sixteen openings are not a list you write to fill that claim.
What they might ask
Deep STAR-method behavioral questions (Nisarg: exact prompts unnamed).
Three named families (6806195), four topic families (8362604), five named (IE.in 2024-grad), three LP names (7623949), two sub-questions (1qj304z), “several” (6425074).
Nisarg Patel — AUTA, Bar Raiser as third live, same day 15:00–16:00 Pacific after 10:00 and 12:30. Candidate: “Deep STAR-method behavioral questions testing alignment to all 16 Leadership Principles.” “Each question explored in detail for ~15–20 minutes.” Exact prompts unnamed. Calendar is one hour. Clash: leave it as a clash. IE-asked as method, not as sixteen interviewer sentences. All-16 claimed on 1/11 locked hours. Reddit 1qj304z: LP-only, ~25 min versus 1 hour. Dump 07 could not re-verify comments this pass; selftext recovered via archive. Exact stems unnamed.
What you say
I’ll take the first Leadership Principle they name and go deep on one project. Situation is short. Action is what I did. Result is one resume number. Then I stop for their why.
They might ask next
If they ask
You answer
Second stem
Second project from the spread (home page / chapter 3). Not a second trophy on the same intern.
Third stem, Dive Deep and Deliver already ran
GiftedBooks, hours to 3–10 minutes.
They keep naming LPs
Backups from the spread. Ylogx primaries max at Ownership, Frugality, Dive Deep. IQVIA at Invent, Learn, Backbone. Argus at Are Right, Highest Standards, Success and Scale. Horizon at Bias, Deliver, Earth’s Best Employer.
Keep straight
Sixteen named openings are not a list you recite to prove coverage.
A backup is a second story, not a fourth primary on Ylogx or IQVIA.
Nisarg exact stems stay unnamed. This Job ID still none.
Tighten
Deep on the stem they named. Stop after one number. Chapter 3 has the sixteen spoken STARs for rehearsal, not for dumping in one hour.
LP default pair
When they do not name a principle, leftover writing still used Dive Deep and Deliver Results as the pair to keep warm. A third, if they want another, is customer time-to-answer on GiftedBooks. Named stems in chapter 3 and 4 override this default. One story, one project, one number family. Two spoken blocks sit here because they are two projects.
What they might ask
Unnamed two-LP / “tell me a time you went deep” / “tell me about delivering.”
Couple LP unnamed “formality” after DSA (LC 7406809).
Named as names only: Deep Dive, Learn & Be Curious, Earn Trust (LC 7623949) — then map names. Full interviewer wording on those names stays unnamed.
Merge lock for unnamed two-LP / all-16: method, not invented interviewer sentences. LC 7406809 couple LP unnamed “formality.” LC 6425074 ~40 min STAR unnamed. LC 7623949 named Deep Dive, Learn & Be Curious, Earn Trust as names. IE-asked as unnamed method. Stories below are Resume-derived.
What you say
Dive Deep — Situation. At Ylogx the chatbot returned SQL and rows, so it looked done.
Task. Every natural-language turn still hit Postgres, and a JOIN could cross organizations if isolation lived only in the app.
Action. I measured bot database latency, put three-tier row-level security and RBAC in Postgres, and put Redis in front with the tenant tier in the cache key so a cached answer could not leak.
Deliver Results — Situation. On Team Horizon the European Rover Challenge date did not move.
Task. A healthy camera feed with a stale occupancy grid is not a delivered rover.
Action. I sequenced GStreamer at 60 frames per second and the live costmap first. Extra hardware was the cut, not the occupancy path.
Result. We placed 17th of more than eighty teams.
They might ask next
If they ask
You answer
Intro already used the rover, they want Deliver
Ylogx 40 percent faster reports and 99.9 percent (Deliver backup). RLS was not in the database yet if they ask what was still open.
Third stem
GiftedBooks, hours to 3–10 minutes.
Feedback / mistake / failed
Argus 73 percent mAP was not shippable for floor alerts; gate 89 percent mAP and 24 FPS, then 20+ cameras.
Conflict / disagreement
IQVIA hybrid versus vector-only, commit on LangSmith evals. No manager quote.
7623949 names after Why Amazon
Dive Deep = this Ylogx block. Learn = IQVIA LangGraph. Earn Trust = GiftedBooks grounded RAG + 99.5 percent.
Keep straight
Dive Deep lead is Ylogx Redis and RLS together. SEO 403 is not that story.
+65, 40 percent, 30 KPI, and sub-210 are other Ylogx families. Dive Deep result here is −35 percent bot DB plus 99.9 percent uptime.
60 FPS is the rover camera pipeline. 24 FPS is the Argus detector budget. They do not share a STAR.
Horizon 17th is we. RLS in Postgres is I. Missed ERC and missed 99.9 percent are not on the resume.
GiftedBooks stays off the Ylogx block.
Tighten
Speak one block, then wait. If they already heard Ylogx, Dive Deep backup is IQVIA ranking contradiction — not a fourth Ylogx trophy. Full sixteen STARs are chapter 3.
Same-story probes
Leftover hours that went well on paper still grilled the same story: why this decision, was that actually challenging, what would you do differently, what did you do versus the team. Restarting Situation wastes the probe. Answer the probe on the path you already opened.
What they might ask
Follow-ups on all three LPs (LC 6806195).
Whether the first story was actually challenging; rate limiting / geofencing / trade-offs as story probes, not LLD (LC 8362604, first 15–20 minutes intense).
Deep cross-questioning on every answer; ~37 min; “giving you back your 20 minutes” (IE.in 2024-grad).
Every metric and decision; candidate: wasn’t about right or wrong (LC 7981646 BR, selected).
Don’t create fake LP answers (LC 6653463).
IE-asked probe style. 10× as interviewer wording: not verified (dump 14 n=0).
What you say
On that same Ylogx path: the JOIN was legal SQL and illegal for the tenant, so the database role had to deny it. Redis without the tier in the key would have cached the leak. What I would sequence earlier is denied-tier tests beside the RLS migration, not a bigger cache first.
They might ask next
If they ask
You answer
Challenging enough (IQVIA was the Invent story)
Clause-ID miss / span-less test on that same IQVIA path. Not a new org.
What would you do differently
Argus → per-camera error budget as future work. IQVIA → freeze gold questions and citation-required writer earlier. Horizon → name the ship-blocker in week one.
What did you do versus the team
I for RLS, Redis, LangGraph nodes, YOLOv9 gate, GStreamer/costmap software. We for ERC 17th and for site −50 percent / 2× after alerts.
Keep straight
A manager fight is not on the resume. Conflict stays hybrid versus vector or RLS-in-database versus app-only WHERE.
Rate Limiter and geofencing are not intern evidence. 8362604 used those words as story probes. Rate Limiter SDE I live count stays 1.
Ylogx primaries are Ownership, Frugality, Dive Deep. A fourth Ylogx LP to fill unnamed probes does not belong.
Tighten
Answer the why on the same path. A fake STAR falls apart on the second why — stay on mechanisms you shipped. Then wait.
If they paste code, Java
A leftover hour can still be a coding hour, including when the label is Bar Raiser or Hiring Manager. Live Code is Java. Restate, pick a structure, say time and space, then type. After a complete answer, ask whether they want a dry-run or a second problem, then wait.
What they might ask
Nice Subarrays (6806195, slug linked by the candidate).
Tree path A/B/C (7406809, unnamed).
Unique Paths-like max points (7035289, unnamed; 62 is not that title).
Distance K, no parent map (7623949 Fluency, unnamed; 863 is not that title).
Union-Find (6866029, unnamed title).
Maximum Sum Linked List common nodes (6881058, unnamed id; 962/210 are not that title).
Combine Garlands / Anagrams variation (8362604 HM, not BR).
IE-asked as families. Standard CS for the Java approach. Not a list for this Job ID. Ids only if the candidate linked them; else unnamed.
What you say
I want to confirm input size, what to return, and whether in-place is required. Brute is all pairs / all subarrays and that is too slow for n. I’ll use this structure, O(n) time and O(n) extra unless you banned extra space. Then I’ll type.
They might ask next
If they ask
You answer
Walk the hot loop
Narrate it. Dry-run one example.
Edges
Name one unprompted: empty, k=0, src equals target, overflow → long.
Does that match what you wanted
Ask that, then stop. Deeper on X only if they say so.
A second problem
Reset constraints. New function. The first one stays as it compiled.
Keep straight
Editor language is Java. Internship production was Python / TypeScript / ROS2.
LP-only on the invite still leaves Java in play (7406809, 6806195).
Unnamed prompts stay unnamed. Guessed LeetCode numbers stay off the title.
Bookstore word-count is LLD, not leftover DSA (Shiwangi earlier live).
Login Tracker is unverified. Rate Limiter live count is 1, not UTA-default leftover LLD.
Tighten
Approach in one minute, then type. Chapter 6 has the families. If they never paste, stay on stories.
Fluency may be a tree
If the invite says GenAI Fluency and they paste a tree, the tree is the work. Write the Java first. One sentence that this is a walk on a tree, not retrieval. RAG talk waits until they still ask after the code.
What they might ask
Nodes at distance K, expected O(n), no parent mapping (LC 7623949 Fluency).
Next Greater (7724048), dynamic k-th (7850431), LCA (8014509) on mixed hours.
How you use GenAI — only if they still ask after the code.
LC 7623949 — labeled Gen AI Fluency; candidate: essentially a DSA round; similar to nodes at distance K; expected O(n); no parent mapping. GenAI prompt not named on that hour. Later BR ~22 min LP-only. IE-asked. LC 7584467: “just another DSA round.” Intern Fluency LC 7643209 min-heap is a different loop — not Distance K, not this leftover default.
What you say
This is a tree. I’ll connect each node to its parent through the child pointers, then BFS from the target until distance K, and collect those nodes. I am not storing a parent field on the class if you don’t want extra mapping — I’ll build an adjacency list in one pass instead. Time O(n), space O(n).
They might ask next
If they ask
You answer
How you use GenAI after the code
I draft with a model; I verify with traces, tests, or a required overlapping span. I do not ship a fluent paragraph with no citation. Then stop.
They never ask GenAI
Leave it.
Keep straight
The Fluency hour on 7623949 stays unnamed as an LC id. 863 is not that title.
Intern min-heap Fluency (7643209) is not this leftover default.
IQVIA LangGraph is not the answer to a binary-tree prompt.
Tighten
Write the tree first. Chapter 7 if they still ask after the code. Empty editor means no LangGraph lecture.
Why Amazon
Why this company, why SDE I, why leave internship work. Judgment under a clock, not childhood worship and not a fake team name. This is a method answer, not a STAR with a trophy. A personal org/team reason is still yours to lock; this page does not mint one.
What they might ask
Why Amazon (LC 7623949 leftover BR: Introduction; Why Amazon; Deep Dive; Learn & Be Curious; Earn Trust as names).
Why SDE I / why not stay at the intern role.
Why Java on this loop.
IE-asked on 7623949. Prince-style loops (Job 3057703, NMF for this ID) also closed on internships plus LP. IGotAnOffer lists Why Amazon as a common opener — Class C as a guarantee, useful as shape.
What you say
I’m moving from internships into SDE I, where you own a slice and the clock is real. At IQVIA I learned to stop a fluent answer that has no overlapping source span. At Ylogx I learned a SQL chatbot is not done until the database role protects the rows. I want that ownership-and-depth bar at SDE I scope.
Intern scope versus SDE I ownership. No insult to IQVIA.
Why Java on the loop
Live Code is Java. I close the gap in the editor. Internship bullets are not JVM internships.
Keep straight
No team, org, or childhood Amazon story on the resume. None in this answer.
IQVIA stays a place you learned the span-stop. It is not a company you talk down.
You are a candidate in a leftover hour, not a Bar Raiser.
An AWS bill is not on the resume.
Tighten
Three sentences, then wait. If they name an LP after this, use the default pair or the named map — do not retell Why Amazon as a STAR.
Questions to ask them
Near the end they will offer you time. Ask about the work. Do not pitch a product. Compensation belongs in a later recruiter conversation, not this hour.
What they might ask
Do you have any questions for me?
After two LP sub-questions, the same close (Reddit 1qj304z).
Reddit 1qj304z: after two LP sub-questions the interviewer asked if the candidate had questions; the candidate pitched a Prime Video rating idea; the interviewer would not discuss it; the hour ended ~25 minutes. That reverse-Q is a warning, not a template. IE.in 2024-grad interviewer “giving you back your 20 minutes” after enough notes. IE-asked as end-of-hour shape.
What you say
What does a typical week look like on this team for a new SDE I?
What separates someone doing well at six months from someone who is not?
What have you seen new grads struggle with on Live Code versus in production?
What has kept you at Amazon?
They might ask next
If they ask
You answer
They answer with a team problem
One follow-up on on-call or on how SDE I code review works. Then stop.
They say they cannot share
Thank them and stop.
Time remains and they open an editor
Switch to Java. Shiwangi’s leftover DSA existed because time remained.
Keep straight
Compensation, levels, and competing offers wait for the recruiter conversation.
“What does Amazon do” is not a question for this hour.
Prime Video, Alexa, or a startup pitch is not the close (1qj304z reverse-Q ended the hour).
Whether this was the Bar Raiser is not a question you ask.
Tighten
One or two questions from the list, listen, stop. If they open an editor, chapter 6.
If they end early
Several leftover hours ended under the calendar hour. That can mean they have enough notes. It is not a script that you failed, and it is not a script that you passed. You ask your questions, you thank them, you leave the editor in a compiling state if you coded.
What they might ask
That’s all I needed / giving you back your 20 minutes.
Remaining minutes for another story or for code.
Ended early in the hour lock: 7623949 ~22 min; 1qj304z 25 min; 7563011 ~25/35; IE.in ~37 plus “20 minutes back”; 6425074 40 min versus 1 hour. Used full hour: 8362604 BR; 6806195 ~60. Nisarg calendar 1 hour plus all-16 wording. Divanshu minutes unstated. n=11, not a population. IE-asked as timing evidence.
What you say
Thanks for the time. If it is useful I can send a follow-up on the edge case we discussed. Otherwise I’ll wait for the recruiter.
They might ask next
If they ask
You answer
Remaining minutes for code
Take it. Java. Restate constraints.
Remaining minutes for another LP
Next locked story. Do not restart the first.
After the call, what the hour “was”
You do not classify it for the recruiter as a fact you inferred.
Keep straight
Leftover silence is not a sixth project. StratifyLabs / Argus / GiftedBooks wait until they ask.
A live verdict is not something you request.
The slot type stays unlabeled unless they named it. You do not announce Bar Raiser to the recruiter as your inference.
Tighten
Questions from the previous card, thank you, stop. After the call, wait for the recruiter. This booklet still has no Job 10454435 named question.