首次求职示例 · 已冻结的系统输出
一个课程项目，一段实习：这份简历能说明什么？
这位虚构的应届学生用 Python 清理过问卷数据，也协助过每周报表工作。这里展示系统针对一个初级数据分析职位给出的实际结果。
免费扫描我的简历无需账户。
学生、学校、实习团队和职位均为虚构。系统输出来自一次真实的线上免费扫描请求，说明文字由我们撰写。
此案例延续首页虚构的问卷数据清理项目，下方输出来自另一次实际记录的扫描。
三种语言页面均原样展示同一次英文学生案例结果。此学生案例尚未运行中文版本。
捕获时间：2026-09-13T11:14:58Z · 模型：deepseek-v4-flash · 编号：FL-20260913-92a02185。这是一次已记录的结果，并非实时扫描。
再次运行时，措辞或判断可能不同。
模型名称来自该次部署的配置，响应本身未予确认。
此次运行的结构化覆盖表不可用，因此改为展示完整的系统文本与分数。
学生案例：准备申请第一份分析岗位
- 目标：新加坡初级数据分析职位
- 课程项目：负责 Python 数据清理，图表由队友制作。
- 实习：在主管复核下更新 Excel 跟踪表。
查看实际发送的完整简历与职位描述
虚构简历输入 · 英文
Synthetic student S01
Location: Singapore
Target: Junior Data Analyst
Education
BSc Business Analytics, Synthetic University
September 2023 - expected May 2027
Final-year student. The degree is not yet completed.
Course project
Survey data cleaning, January - April 2026
- Cleaned survey responses in Python for a course project.
- Used pandas to remove duplicate rows, standardize response categories, and flag missing values.
- Recorded the cleaning steps in a notebook and explained the choices in a README.
- My contribution was data cleaning. A teammate produced the charts.
- This was coursework, not client work or a deployed analytics platform.
Internship
Operations intern, Synthetic Operations Team
June - August 2026
- Helped with weekly reporting.
- Updated an Excel tracker and checked entries for missing dates under a supervisor's review.
- Summarized unresolved entries for the supervisor.
- I did not own the reporting process or measure a business improvement.
Skills
Python and pandas used in the course project.
Excel sorting, filtering, and lookup formulas used during the internship.
SQL SELECT, WHERE, and GROUP BY practiced in coursework.
No production-database experience claimed.
All people and organizations in this CV are fictional.
虚构职位描述输入 · 英文
Synthetic vacancy: Junior Data Analyst
Location: Singapore
This is a fictional job description, not a live vacancy.
We welcome final-year students and recent graduates for an entry-level analyst role.
Requirements
- Clean and check tabular data using Python or Excel.
- Write basic SQL queries, including joins.
- Explain data-cleaning choices and document the steps taken.
- Work with a supervisor and communicate issues clearly.
- Handle data within the team's access and confidentiality rules.
Useful additional experience
- Create a simple chart or dashboard and explain what it shows.
Applicants may use coursework or internship examples.
Production-system ownership and prior full-time analyst employment are not required.
已记录的结果
匹配分数
78/100
判断
Apply
最有力的切入点
The candidate can defend a full data-cleaning workflow end to end: pandas dedupe, category standardization, missing-value flagging, a notebook of steps, and a README explaining the choices, which is precisely the JD's core cleaning-and-documentation requirement.
系统给出的示例改写
原文
Used pandas to remove duplicate rows, standardize response categories, and flag missing values.
改写后
Cleaned survey responses in Python with pandas: removed duplicate rows, standardized response categories, and flagged missing values, then recorded each step in a notebook and explained the choices in a README.
The rewrite keeps every action the source states and pulls the documentation step (already in the CV) into the same bullet so it answers the JD's "explain data-cleaning choices and document the steps" requirement in one line. It adds no outcome, no scale, and no tooling beyond what the source names. If the candidate can state the number of responses cleaned or the categories standardized, that would let the bullet say more, but only with the actual figure.
待补证据
No evidence of SQL joins, which the JD names explicitly alongside basic queries.
No chart or dashboard produced by the candidate; the CV states a teammate produced the charts.
No stated familiarity with the team's access or confidentiality rules, only supervised work.
No artifact named for the internship tracker (no sample, no schema, no size).
风险表述
This is the interview attack surface, what a careful reviewer is most likely to push back on, not a list of mistakes. Read it as preparation.
"SQL SELECT, WHERE, and GROUP BY practiced in coursework."
- Why an experienced reviewer would push back: the JD asks for joins, and the CV lists three clauses without joins, so a reviewer will test whether the candidate can actually write a join or has only seen one demonstrated.
- One concrete artifact to bring: a short notebook or .sql file from coursework showing at least one join the candidate wrote, with the two tables and the join key visible.
"Helped with weekly reporting" during the June to August 2026 internship.
- Why an experienced reviewer would push back: "helped with" is vague on what the candidate personally did versus observed, and the JD wants clear communication of issues to a supervisor, which needs a concrete instance.
- One concrete artifact to bring: a redacted copy of one weekly tracker update plus the summary of unresolved entries the candidate sent to the supervisor.
"Updated an Excel tracker and checked entries for missing dates under a supervisor's review."
- Why an experienced reviewer would push back: the scope is narrow (missing dates only) and the reviewer cannot tell the tracker's size, the error rate, or whether the candidate found anything, so the line reads as task-following rather than analysis.
- One concrete artifact to bring: a redacted before/after view of the tracker showing the entries flagged and the supervisor's review notes.
下一步
Apply, and before submitting, write one small SQL exercise that uses a join on two tables and keep the file, so the SQL line can name joins honestly rather than only SELECT, WHERE, and GROUP BY.
结果摘要
The CV maps cleanly onto the core of this JD: the course project states the candidate personally cleaned survey responses in Python with pandas (dedupe, category standardization, missing-value flagging), recorded the steps in a notebook, and explained the choices in a README, which is exactly the "clean and check tabular data" plus "explain data-cleaning choices and document the steps" pair. The internship adds supervised Excel tracker work and summarizing unresolved entries to a supervisor, which matches the "work with a supervisor and communicate issues" requirement. The single biggest factor keeping the score from being higher is SQL: the CV only claims SELECT, WHERE, and GROUP BY practiced in coursework, and the JD explicitly asks for joins, which are not named anywhere in the evidence.
分数明细
| 评分项目 | 得分 | 满分 | 说明 |
|---|---|---|---|
| 硬性要求覆盖 | 27 | 35 | Python/pandas cleaning, documentation, supervisor communication, and confidentiality-adjacent supervised work are directly stated; SQL joins are the one named requirement with no evidence. |
| 经验相关度最大短板 | 22 | 30 | The course project and the operations internship are both entry-level and directly in the data-cleaning and supervised-reporting lane the JD describes, though neither is analyst-titled. |
| 证据质量 | 17 | 20 | The CV is unusually specific about scope and limits (notebook, README, "my contribution was data cleaning", "I did not own the reporting process"), which is strong for a student. |
| 陈述可辩护度 | 12 | 15 | Almost every line is already hedged to what the candidate can defend; the only soft spot is the SQL skill line, which names clauses but not joins. |
分数与解释来自本次捕获，评分项目名称属于界面文字。
压力测试：无法自洽的交易员声明
我们也保留这项较早的虚构测试：在简历中植入一条与大学就读时间冲突的“2008 年领导交易台”声明，检查扫描是否会提出质疑。
这是 2026 年 9 月的历史结果，保留当时实际返回的内容，不表示当前系统会给出完全相同的结果。
交易员案例分别运行了英文、简体中文和繁体中文版本。结果存在差异，详见下方说明。
同一语言的独立重跑结果稳定：英语为 42/42，简体中文为 28/28。这次捕获也如实显示了跨语言差异：英语返回 42「Needs more evidence」，两次中文运行均返回 28「Skip」。
2026-09-04T06:13:43Z 用 deepseek-v4-flash 跑出，order FL-20260904-af58d37b。这是冻结的输出：同一份 CV 再跑一次，措辞会不一样。
查看保留的交易员压力测试
已记录的结果
匹配分数
28/100
判断
Skip
最有力的切入点
候选人作为 FX 交易员对 G10 和亚洲 FX 现货及远期的定价经验，与 JD 要求的 FX 产品理解直接匹配，这是唯一可清晰辩护的交叉点。
Fact-Checker 拒绝改写被标记的表述
The lines that need work here are the ones flagged above; they need removing or evidence, not rewording.
待补证据
任何独立主导的量化研究项目，从假设提出到回测、再到交易台反馈的完整闭环记录。
可展示的 Python 建模作品，例如信号测试脚本、回测框架或数据分析 notebook，而非仅描述为"built scripts"。
发布或内部发表的研究文档、备忘录或交易台可执行的研究笔记样本。
与工程师合作将原型推向生产环境的具体案例，而非"reviewed model assumptions with technology"这类模糊表述。
模型性能衰减分析或信号有效性追踪的量化结果。
风险表述
3 lines·2 need proof·1 cut
这是面试攻击面，即审慎的评审者最可能质疑的地方，而非错误清单。候选人应将其视为准备材料。
LED THE ENTIRE LONDON FX TRADING DESK DURING THE 2008 FINANCIAL CRISIS, MANAGING A USD 50 BILLION BOOK AND DELIVERING USD 300 MILLION P&L IN THE FIRST QUARTER.
CUT- 为什么经验丰富的评审者会质疑：CV 显示候选人 2008-2011 年在 University of Bristol 攻读 BSc Economics 全日制本科，该声明要求候选人在本科第一年同时领导伦敦整个 FX 交易台并管理 500 亿美元账簿，且 CV 中无任何实习、休学或兼职解释可调和这一冲突。此外，该声明以全大写形式突兀出现在经历列表末尾，无雇主名称、无时间范围，与前后格式完全脱节。
- 候选人应携带的具体证据：若此声明属实，需提供 2008 年 Q1 的雇佣合同、雇主出具的职位证明信、以及经审计的 P&L 报告；若无此类文件，该行应从 CV 中删除。
FX trader, StraitsBridge Bank, 2015-2021 与 Senior trader, StraitsBridge Bank, 2021-2026 之间的晋升路径。
NEEDS PROOF- 为什么经验丰富的评审者会质疑：2021-2026 的时间范围意味着该职位描述到 2026 年，若当前为 2025 年或更早，则包含未来日期；若当前为 2026 年之后，则需确认离职状态。时间范围模糊且无明确结束原因。
- 候选人应携带的具体证据：雇佣记录或离职证明，明确各职位的起止月份。
"partnered with quants on backtesting carry and momentum screens" 和 "reviewed model assumptions with technology"。
NEEDS PROOF- 为什么经验丰富的评审者会质疑：这些措辞表明候选人处于辅助或咨询角色，而非 JD 要求的"owning a model lifecycle from hypothesis to desk feedback"。评审者会质疑候选人是否真正编写过回测代码、是否独立提出过信号假设，还是仅作为交易员提供市场直觉。
- 候选人应携带的具体证据：其参与过的回测项目的代码提交记录、会议纪要以明确其具体贡献范围，或交易台实际使用过的研究笔记样本。
下一步
在删除或补充 2008 年声明证据之前，不建议申请该量化研究岗位；若候选人希望转向量化方向，应先积累可展示的独立研究项目，再考虑申请量化研究或量化开发类初级岗位。
结果摘要
该分数基于以下判断：JD 要求的是具备完整模型生命周期所有权、可发布可执行研究的量化研究员，而候选人证据显示其核心身份是交易员，研究相关经验均为辅助性或合作性质，且存在一条无法自洽的极端时间线声明。分数主要受限于候选人缺乏独立主导量化研究项目的直接证据，以及一条严重损害整体可信度的经历声明。
分数明细
| 评分项目 | 得分 | 满分 | 说明 |
|---|---|---|---|
| 硬性要求覆盖最大短板 | 8 | 35 | JD 要求 Python/SQL 建模、FX/rates 产品理解、发布可执行研究、拥有模型生命周期，候选人仅有 FX 交易经验和辅助性脚本工作，无独立研究发布或生命周期所有权证据。 |
| 经验相关度 | 10 | 30 | 交易经验与 FX 产品直接相关，但研究、建模、与工程师合作均停留在"partnered with"和"reviewed"层面，非 JD 要求的 hands-on research 角色。 |
| 证据质量 | 6 | 20 | 多数经历描述宽泛，无具体研究产出、模型指标、代码仓库或可验证的分析成果；唯一带数字的经历声明与时间线严重冲突。 |
| 陈述可辩护度 | 4 | 15 | 2008 年金融危机期间领导伦敦交易台的声明与 CV 中 2008-2011 年本科在读的时间线直接矛盾，且无任何可调和解释，该声明无法在面试中站住脚。 |
分数与解释来自本次捕获，评分项目名称属于界面文字。