Claude prompts for NLP and LLM engineer job searches

I am applying for the NLP / LLM engineer role below. Rewrite my resume to match it using only experience already on my resume. Do not invent models, datasets, frameworks or benchmark numbers. The posting may ask for things like fine-tuning, retrieval-augmented generation, evaluation, prompt engineering, tokenization or serving; for each requirement, find the closest real evidence on my resume, and if there is none, put it under "Gaps" instead of writing it in. Output: a table with the columns requirement, evidence, rewritten bullet; a one-page revised resume in plain text where each bullet names the task, the approach and the result; and the Gaps list with one honest sentence per gap about what I have done that is adjacent. Keep all jobs, titles and dates unchanged.

Job description:
[paste job description]

My resume:
[paste your resume]

Write a cover note of at most 120 words for the NLP / LLM engineer role below. Use only facts from my resume and my notes. No invented results, no "passionate", no "cutting edge". Structure: one sentence that shows I understand the language problem this team is working on (take it from the job description), two sentences on the most relevant thing I have built with text or language models and what it did, with a number if I have one, and one closing sentence with a specific ask. Then give a 50-word version for a LinkedIn message. Sentence case, plain words, no exclamation marks.

Job description:
[paste job description]

My resume:
[paste your resume]

Why this team, in my words:
[two or three lines]

Act as the interviewer for a 45-minute NLP / LLM engineer technical screen at the company below. Ask one question at a time and wait for my answer. Cover, in order: one Python coding question on text processing (for example tokenizing and counting n-grams, or implementing a simple retrieval scorer), two fundamentals questions drawn from the job description (for example how attention works, what tokenization choices affect, the difference between fine-tuning and instruction tuning, why retrieval-augmented generation fails, or how to evaluate a summarizer), one question about a project on my resume, and one short question about the cost or latency of serving a model. After each answer, score it 1 to 5, give a strong answer, and name the one thing I should have said. End with three topics to review tonight.

Job description:
[paste job description]

My resume:
[paste your resume]

Run an LLM system design interview with me for the role below. Pick a problem that fits the company (for example customer support answers over internal documents, structured extraction from contracts, or search with generated summaries) and describe it in two sentences. Guide me one stage at a time, waiting for my answer: requirements and what "good" means for the user, choosing between prompting, retrieval and fine-tuning, the data and how I would build an eval set, chunking and retrieval design if relevant, prompt and output structure, guardrails for wrong or unsafe answers, latency and cost per request, caching, monitoring in production, and how I would iterate. After each stage, point out what an experienced engineer would have raised that I missed. Finish with a one-page summary of the design in my corrected words. Do not give me the full answer up front.

Job description:
[paste job description]

What I know about the product:
[a few lines]

Help me prepare behavioral answers for an NLP / LLM engineer interview using only the situations I describe below. Do not add details; if a story is missing an outcome or a number, ask me for it. First list the eight questions most likely for this role and company (a model that produced wrong or harmful output, a disagreement about evaluation, shipping under a deadline, a stakeholder who wanted an LLM where a rule would do, a mistake you owned, learning something new quickly, mentoring, and a project that changed direction). Then map each to one of my stories and write an answer of 150 to 200 words in the shape situation, action, result, what I would do differently. Flag where the same story is used twice so I can spread them out.

Job description:
[paste job description]

My stories, rough notes are fine:
[three to five real situations]

Turn my notes below into a portfolio write-up of 500 to 700 words about an NLP or LLM project, first person, plain language, aimed at a hiring manager. Use only what is in my notes. Where something important is missing (the task definition, dataset size and source, the baseline, the eval method and its numbers, cost or latency, failure cases), insert a placeholder in square brackets and list the placeholders at the end. Structure: the problem and who it was for, the data and its problems, the approach and what I tried first that did not work, how I evaluated it and what the numbers say and do not say, example failures, and what I would do next. Add a suggested title, a two-sentence summary for the top of a README, and three honest resume bullets.

Project notes:
[paste notes, commit messages or a rough description]

I have an offer for an NLP / LLM engineer role. Below are the offer details, my current compensation, any competing offers and what I have found from public sources. Help me evaluate and negotiate honestly; do not invent market data, and tell me what to look up if you need a number I have not given. Output: (1) a table of four-year total compensation under a cautious and an optimistic equity assumption; (2) which components are most negotiable for this kind of company and why; (3) a short script for the call that asks for a specific number with a reason tied to my experience, in a collaborative tone; (4) questions to ask before accepting, including the level, vesting, refreshers, compute and data access for the team, on-call, and whether the role is research-leaning or product-leaning. Under 500 words.

Offer:
[base, bonus, equity and vesting, signing bonus, location]

My situation:
[current compensation, competing offers, constraints]

Write a 30/60/90-day plan for me as a new NLP / LLM engineer on the team below. Keep it realistic: I have to learn the codebase, the data, the eval setup and the people before shipping. Days 1 to 30: which models and pipelines are in production, how they are evaluated, which datasets exist and who owns them, and what the top failure cases are. Days 31 to 60: a first contribution, for example an eval gap closed, a prompt or retrieval fix with a measured effect, or a cost reduction. Days 61 to 90: a measurable project with a success metric and risks. For each block, give three to five outcomes rather than activities and one question to ask my manager in week one. Mark assumptions about the team that I should verify. Under 450 words.

Job description:
[paste job description]

What I learned about the team in interviews:
[notes]

Help me design an evaluation set for the LLM task described below, the way I would present it in an interview or a take-home. Ask me first for anything you need: the task, the users, what a wrong answer costs, and what data I can use. Then produce: (1) the dimensions to evaluate (for example correctness, groundedness against sources, format compliance, refusal behavior, latency) with a one-line definition of each; (2) a plan for 100 to 300 examples covering normal cases, edge cases, adversarial inputs and known failure modes, with rough proportions; (3) for each dimension, whether to score with exact match, a rubric, a model grader or a human, and the risk of each choice; (4) how to check that the grader itself agrees with humans; (5) what I would track over time in production. Keep it under 600 words and say what I should not claim about a result from a set this size.

The task:
[describe the feature or take-home]