資料科學家的英文履歷怎麼寫
更新於 2026-09-02

資料科學家的英文履歷,計量單位是上線後改變了某個數字的模型。Kaggle 名次、[N]% 準確率、論文清單,在招募主管眼裡都是過程。他在找三件事:哪個模型進了 production、它取代了什麼決定、之後哪個指標動了多少。
這一頁和資料分析師、機器學習工程師各差一個字。分析師的證據是被分析改變的決策,機器學習工程師的證據是模型服務的延遲與可用性。資料科學家的證據在中間:一個你建的模型,接手了原本靠規則或人判斷的決定,而且上線之後有人替它守著監控。「Built a churn model with [N]% accuracy」缺的正是後半句。
範例全部是英文,因為外商履歷的成品是英文。數字全部留 [N],因為面試時要替那些數字辯護的人是你。以下依序是版面順序、三個層級的 summary、依職能分組的 bullet、改寫對照、技能取捨、ATS 關鍵字、常見錯誤與面試追問。
這個職位的英文履歷,完成後長這樣
下面是一份虛構人物的完整範例,由產品的排版引擎渲染成真正的頁面,分數由同一套規則引擎現算——不是我們打上去的。缺的數字一律留 [N],所以分數停在上限以下;那不是範例的缺陷,是產品的規則在示範。
Noor Siddiqui
Education
- Built a hierarchical demand forecast for a metropolitan transit network as the thesis project, beating the agency’s seasonal baseline by [N] points of error
- Won the department’s applied statistics thesis prize for that forecast, which the transit agency adopted into its quarterly service planning cycle
Experience
- Built the demand forecast that sets weekly driver incentives across [N] cities and [N] million rides a month, cutting forecast error by [N] points and incentive spend by [$N] a quarter
- Halved rider cancellations on surge trips by replacing the flat surge multiplier with an accept-probability model, launched after a [N]-city holdout held the gain for [N] weeks
- Designed the switchback experiment framework for marketplace pricing, so two-sided interference stopped inflating lift; [N] pricing changes shipped on its readouts in one year
- Owned the promo-abuse model that scores every new account, catching [N] thousand fraudulent signups in its first quarter at a false-positive rate the support team agreed to in advance
- Developed the claims severity model that replaced the actuarial lookup table for auto claims, cutting reserve error by [N]% across the open claims book
- Deployed a churn score to the renewals team as a weekly file with a one-page playbook, lifting retained premium by [$N] in the first renewal cycle
- Introduced uplift modelling for retention discounts, so offers went only to policyholders whose renewal the discount changed; cut offer spend by [N]% at an unchanged retention rate
- Established calibration and drift monitoring on the team’s [N] production models, catching a feature outage that had silently degraded the fraud score for [N] days
Projects
- Created an open-source rolling-origin backtesting harness for time-series forecasts, adopted by [N] teams and downloaded [N] thousand times
Additional Information
版面順序:先讓人看到上線的模型
單欄、標準標題、不放模型架構圖,也不放 ROC 曲線。第一次掃描的時間讀不完一張圖,解析器則連讀都不讀。
順序的原則只有一個:進了 production、改變過數字的模型先出現。學術出身的人習慣把論文與學位放最前面,外商的招募主管會直接翻到 Experience 找模型的下落。
- 1Summary
三句話:年資與領域、一個上線後改變了指標的模型、你常用的方法族(預測、因果、排序)。看的人在這裡決定你是建模型的人,還是跑過 notebook 的人。
- 2Experience
倒序。每條 bullet 一個模型或一個實驗:它取代了什麼、上線在哪、哪個數字動了。方法名稱放句中,不放句首。
- 3Projects/Publications(選用)
博士與轉職者放在 Experience 之後。論文只列與職缺相關的一到兩篇並附連結;Kaggle 只在前 [N] 名、而且講得出自己那段方法時才寫。
- 4Skills
按用途分組:建模、實驗與因果、上線與監控、資料與工具。只列你能在白板上推導、在面試裡調參的。
- 5Education
碩博士是這個職位的常態,三行寫完:學位與領域、一項相關的研究產出,以及那項產出後來被誰採用。GPA 畢業三年後刪掉。
三個層級的 Summary 範例
Summary 預告的是模型的下落。三行之內要出現一個進了 production 的模型,以及它移動的那個數字。
Statistics M.Sc. with two data science internships, focused on models that reach production. Built the demand forecast now scheduling weekly stock for [N] stores of a grocery chain, cutting stock-outs by [N]% against the buyer’s manual forecast. Seeking a data scientist role where the person who builds a model also watches it after launch.
新鮮人的證據是一個有人在用的模型,加上它打敗了誰的基準。最後一句說明你知道上線之後還有事要做。
Data scientist with [N] years across ride-hailing and insurance, specialising in forecasting, uplift modelling and experiment design. Owns [N] models in production, including the surge-pricing model that halved cancellations on surge trips and the promo-abuse model that recovered [$N] in its first quarter. Reads out experiments that finance signs off on without a second analysis.
中生代的關鍵字是 owns:模型上線後仍然歸你管。兩個模型各配一個被移動的數字,每個 [N] 都是一題可以追問的事。
Lead data scientist who decides which problems deserve a model: retired [N] models nobody acted on and pointed the team at the [N] decisions that moved margin by [N] points. Set the experimentation standard (pre-registration, power, switchbacks for marketplace tests) that [N] product teams run on. Measures the team by decisions changed, not by models trained.
資深的證據是取捨:砍掉沒人用的模型、定出實驗標準。最後一句替這個職位的成功下了定義,也是整場面試會被檢驗的主張。
依職能分組的 bullet 範例
資料科學家被評價的是這六件事。每條 bullet 的骨架相同:動詞、模型或實驗、它取代的決定、[N] 結果。
預測與規劃
· Built the hierarchical demand forecast that sets weekly driver incentives across [N] cities, cutting forecast error by [N] points and incentive spend by [$N] a quarter.
· Replaced the finance team’s spreadsheet revenue forecast with a gradient-boosted model backtested over [N] quarters, narrowing the guidance range given to the board by [N]%.
· Cut stock-outs at [N] stores by [N]% with a promotion-aware demand model, after showing that the incumbent forecast missed [N] of the year’s largest promotional weeks.
實驗設計與判讀
· Designed the switchback experiment framework for marketplace pricing, so two-sided interference stopped inflating lift estimates; [N] pricing changes shipped on its readouts in one year.
· Stopped a launch that showed a [N]% lift by tracing it to a novelty effect gone by week three, then wrote the pre-registration template the team has used since.
· Introduced sequential testing with fixed error rates, cutting median experiment duration from [N] weeks to [N] without raising the false-positive rate above the pre-registered level.
因果推論與 uplift
· Built an uplift model for retention discounts, targeting only policyholders whose renewal the offer changed; cut discount spend by [N]% at an unchanged retention rate.
· Estimated the causal effect of same-day delivery on repeat purchase with a difference-in-differences design across [N] staggered city launches, reversing a plan to expand it.
· Replaced last-click attribution with an incrementality model calibrated on [N] geo holdouts, moving [$N] of paid budget out of channels that showed no lift.
風險與詐欺模型
· Owned the promo-abuse model that scores every new account, catching [N] thousand fraudulent signups in its first quarter at a false-positive rate the support team could absorb.
· Developed the claims severity model that replaced the actuarial lookup table for auto claims, cutting reserve error by [N]% across the open claims book.
· Cut credit losses on the pay-later product by [N] points with a default model whose reject-inference step scored the applicants the old scorecard had never seen.
上線與監控
· Deployed a churn score to the renewals team as a weekly file with a one-page playbook, so agents called the [N] highest-risk accounts first; retained premium rose [$N] in the first cycle.
· Established calibration and drift monitoring on all [N] production models, catching a feature outage that had silently degraded the fraud score for [N] days.
· Retired [N] models whose scores nobody acted on, after auditing downstream usage, and moved the team’s time to the [N] that still changed a decision.
把模型變成決定
· Turned a lifetime-value model into a bidding rule the growth team runs without a data scientist, raising payback-adjusted return on ad spend by [N]%.
· Wrote the one-page decision memo behind every model launch, stating the baseline, the expected lift, and the metric that would trigger a rollback.
· Convinced the pricing committee to run a [N]-city holdout on the new surge model instead of launching nationally, which showed the cancellation gain held only in the largest cities.
改寫前後對照
弱句與強句寫的是同一段經歷。共同病灶:弱句以模型指標收尾,強句以被移動的業務數字收尾。
Built a churn prediction model with [N]% accuracy using XGBoost.
Built the churn model behind the renewals team’s weekly call list, lifting retained premium by [$N] in its first cycle; the threshold was set against the cost of a missed call, not a benchmark.
差在哪 · 病灶:準確率是模型的指標,不是公司的指標。改寫說出模型被誰用、怎麼用、哪個數字動了,順便說明閾值是怎麼選的。
Worked on machine learning projects for the marketing team.
Developed the uplift model that decides which customers receive a retention discount, cutting discount spend by [N]% at the same retention rate.
差在哪 · 病灶:Worked on 只證明出席,projects 沒有數量也沒有下落。改寫指名一個模型、它做的決定,以及守住的數字。
Applied advanced statistical techniques to analyze large datasets.
Estimated the causal effect of free delivery on repeat purchase with a difference-in-differences design across [N] city launches, reversing a plan to expand the programme.
差在哪 · 病灶:advanced 與 large 都沒有量,technique 沒有名字。改寫給出方法、設計、資料範圍,以及被推翻的那個決定。
Deployed models to production using Docker and Kubernetes.
Deployed the demand forecast into the weekly scheduling run for [N] cities, with drift alerts that caught a feature outage before the schedule shipped.
差在哪 · 病灶:這是機器學習工程師的句子,而且只有工具。資料科學家的上線證據是模型接上了哪個決定、誰在監控。
Ran A/B tests and reported results to stakeholders.
Designed the switchback experiment for marketplace pricing after showing that user-level A/B tests overstated lift by [N] points; the pricing committee now launches only on switchback readouts.
差在哪 · 病灶:跑測試、寫報告是流程,沒有判斷。改寫有一個你發現的方法問題、你的替代設計,以及被改變的決策流程。
技能區:留什麼、刪什麼
每一項技能都是一道面試題。留下你能在白板上推導、在面試裡調參的;其餘刪掉。
值得佔版面的
Python(pandas、scikit-learn、statsmodels) · SQL · gradient boosting(XGBoost、LightGBM) · 時間序列預測 · PyTorch(有上線模型才列)
A/B 與 switchback 實驗設計 · power analysis 與 sequential testing · uplift modelling · difference-in-differences、合成控制
模型監控(drift、calibration) · MLflow · Airflow · feature store(實際用過才列)
BigQuery · Snowflake · Spark(處理過超出單機記憶體的資料才列) · Git
可以刪掉的
- Machine learning/AI(當成一個項目) · 這是職稱,不是技能。寫方法族與模型類型:forecasting、uplift、gradient boosting。
- Deep learning(沒有上線模型) · 課程作業裡的 CNN 撐不過第一個追問。沒有 production 模型就不列,或改寫成 Projects 裡一條講清楚資料與結果的 bullet。
- Kaggle 名次(前段以外) · 名次不是技能。前 [N] 名且講得出方法的比賽寫進 Projects 一行;其餘刪掉。
- Storytelling、business acumen · 形容詞。bullet 寫出模型改變的決定,這件事就不證自明。
ATS 關鍵字
職缺描述反覆出現的字就是系統比對的字。以下依類別列出這個職位常被比對的詞,拼法與職缺原文一致。
只放經歷撐得起的詞。寫了 causal inference,就要答得出你的識別假設是什麼、哪裡可能被違反。寫了 production,就要答得出誰在監控、多久重訓一次。撐不住的詞不是加分,是你自己埋的雷。
履歷之外的證據:論文、preprint、名次,和那一個 repo
這一頁其他段落談的是紙上的句子,這一段談的是句子指向的東西:一個 arXiv 編號、一張公開的排行榜、一個打得開的 repo。對方會去點,而且點開看到的東西必須跟你寫的一致。
- 他先點哪一條連結
業界的資料科學職缺,對方通常先點 header 那條 repo 連結,找一個他讀得懂的專案;論文標題是之後才貼進搜尋框的。研究導向的團隊順序相反。不論哪一種,header 只留兩條:一個 GitHub,一個 Google Scholar 或 arXiv 作者頁,和 email 排在同一行。第三條連結不會被點開,只是多給對方一個失望的入口。
- 一篇論文在履歷上的樣子:一行,附作者序
寫成「Hierarchical demand forecasting under sparse promotions. [Venue], [Year]. arXiv:[ID] — second of four authors; built the model and the backtest.」刊名與編號讓它可以被查,作者序讓對方不必猜。與職缺相關的留一到兩篇,其餘交給 Scholar 那條連結。九個標題、沒有一條連結,讀起來就是一份忘了改成履歷的 CV。
- 「Under review」的誠實寫法
寫「Under review at [Venue], submitted [Month Year]」;不想寫刊名就寫「Manuscript under review」。手上沒有接受信,就不要寫 accepted 或 to appear,也不要用 forthcoming 把兩件事混成一件。被拒的那一週就把它降回 preprint。同一行 under review 跟著你的履歷過了三年,仔細的讀者會發現,而且會連帶不信任它上面那幾行。
- 競賽名次:名次查得到,百分比要有母數
Kaggle 的排行榜是公開且永久的,所以寫查得到的版本:「Kaggle [competition] — placed [N] of [N] teams, username [handle]」。只寫 top [N]% 而不給參賽隊數,對方會自己換算,順便發現你是四人隊伍裡的第四個。是隊伍就寫隊伍,然後說出你負責哪一段:特徵、驗證切法,還是最後的 ensemble。講得出你的 cross-validation 為什麼沒有 overfit public leaderboard,比一面銅牌有用。
- 一個 repo,而且是有 README 的那個
多數資料科學家的 GitHub 是一排沒有說明的 notebook,那種連結不如不放。pin 一個就好:README 前三行講完資料哪裡來、怎麼跑、結果是什麼。履歷上指得精準:「github.com/…/backtest-kit — rolling-origin backtesting harness for time-series forecasts」。如果帳號裡只有 tutorial 的 fork 和一份 Titanic notebook,header 那一欄就留白。留白不扣分;一條點開是課堂作業的連結,會把它上面那一段一起拖下水。
- 模型是公司的、程式碼不能帶走的時候
這是常態,面試官都知道。不要截內部儀表板的圖,也不要把公司程式碼推到個人帳號——那是資安事件,不是作品集。改成寫方法與決定:「Uplift model for retention discounts (proprietary; method and evaluation design available to discuss)」。真的想要一個點得開的東西,就用公開資料把同一個方法重做一次,並說清楚那是重做的版本。
- 寄出去以前,自己把每條連結點一次
用無痕視窗一條一條點:Scholar 頁面可能是私密的,arXiv 作者頁上你的名字可能有第二種拼法,repo 可能還是 private,Kaggle 帳號可能是空的。這一段最常見的失分不是造假,是對不上——履歷寫 published、頁面顯示 preprint;履歷寫個人專案、commit 紀錄只有一筆 initial commit。對不上的時候,改的是履歷那一行。
資料科學家履歷的常見錯誤
- 1把模型指標當成成就
AUC、F1、RMSE 是你選模型時看的數字,招募主管看的是它們換成了哪個業務數字。一條 bullet 只有 [N]% accuracy 而沒有下文,等於承認模型沒有上線。
- 2整份履歷看不出哪個模型進了 production
notebook 裡的模型與排程裡的模型是兩種履歷。每個模型都寫清楚下落:接上哪個決定、多久重訓、誰在監控。真的沒上線,就誠實寫成 prototype,並說明卡在哪。
- 3讀起來像資料分析師
dashboard、SQL 報表、ad-hoc analysis 佔了一半版面,職能訊號就糊了。分析師的成品是決策,你的成品是接手決策的模型。分析工作縮成一到兩行。
- 4讀起來像機器學習工程師
Docker、Kubernetes、latency、serving 是隔壁職缺的關鍵字。資料科學家的上線證據是模型接上了哪個決定,不是它跑在哪個 cluster。工具留一行,篇幅給被移動的數字。
- 5學術履歷原封不動
論文按年份排、方法寫在句首、沒有一個業務數字。外商看研究產出時只問一件事:這個方法解過什麼真實問題。留一到兩篇與職缺相關的,其餘刪掉。
- 6實驗只寫跑了幾次
「Ran [N] A/B tests」是產能,不是判斷。寫你擋下的一次假 lift、你改過的一個設計、被你的判讀改變的一個決定。
每句主張會招來的追問
外商面試官讀資料科學家履歷的方式,是挑一個模型往下鑽三層:怎麼驗證、怎麼上線、上線後誰在看。
Halved cancellations on surge trips with an accept-probability model.
基準是什麼?holdout 怎麼切、切多久?同一時期還有沒有別的改動?效果有沒有隨時間衰減?
基準的定義、holdout 的設計與時長、同期變動的清單,以及上線後三個月的曲線。答不出衰減,這個 halved 會被當成上線第一週的數字。
Cut forecast error by [N] points against the incumbent forecast.
誤差用什麼量?backtest 怎麼切,有沒有 leakage?誤差降了,決策改了什麼?
誤差指標的選擇理由、rolling-origin backtest 的細節,以及誤差降低換成的業務結果:少訂了多少貨、省了多少獎勵金。
Estimated the causal effect with a difference-in-differences design.
平行趨勢的假設你怎麼檢查?處理組是怎麼被選中的?如果我說那是選擇效應,你怎麼回答?
事前趨勢圖、處理組的指派機制、你做過的穩健性檢查,以及一個你承認無法排除的干擾。因果題的可信度來自你主動講出限制。
Established drift monitoring on all production models.
監控的是什麼:輸入分佈、預測分佈,還是實際結果?警報響過幾次?有幾次是真的?
監控的指標與閾值、警報紀錄,以及一次真的抓到問題的事件。從沒響過的監控,面試官會當成裝飾。
Stopped a launch that showed a [N]% lift.
你怎麼發現那是假的?產品團隊反對嗎?你用什麼說服他們?
發現問題的分析步驟、對方的論點、你最後怎麼溝通。這題測的是判斷力,以及你敢不敢說不。
常見問題
我的模型都停在 notebook,沒有真的上線,怎麼寫?
誠實分層。寫成 prototype,並講清楚它驗證到哪一步:backtest 的結果、跟現行方法比的差距、卡在哪個環節。「Validated a churn model against the incumbent rule in a [N]-month backtest; deployment blocked by data-access approval」比假裝上線可信得多。
同時找一件確實改變過決定的事,即使沒有模型:一次實驗判讀、一次被採用的預測。那才是這個職位的計量單位。
博士畢業、只有論文和研究經歷,履歷要怎麼調整?
把研究改寫成問題、方法、結果的格式,並替每一項配一個可查證的產出:程式碼、資料集、被引用或被採用的地方。論文只留一到兩篇與職缺相關的,方法名稱放句中。
Projects 區放一個用真實資料、面向真實決定的專案,公開資料也可以。外商在意的是你能把方法接到問題上,不是你讀過多少方法。
Kaggle 名次要放嗎?
前 [N] 名、而且你講得出自己那一段方法的,放 Projects 一行。其餘刪掉。名次是過程,面試官會問的是你在那個比賽裡做對了什麼、換到真實資料會壞在哪。
我做的模型是公司機密,數字不能寫,怎麼辦?
寫相對變化與範圍:誤差降了 [N] 個百分點、模型覆蓋 [N] 個城市。金額、客戶數、特徵細節都不需要。範例裡的 [N] 就是這個用法:數字由你在安全範圍內自己填。
資料科學家與機器學習工程師的履歷差在哪?我兩邊都投。
同一段經歷寫兩個版本。投資料科學家,bullet 收尾在被移動的業務數字與實驗判讀;投機器學習工程師,收尾在延遲、吞吐、服務可用性與重訓管線。兩份都誠實,只是主詞不同。一份履歷同時寫兩種,通常兩邊都不像。
把你的模型,寫成改變過數字的樣子
用中文照實描述你建過的模型、它上線在哪、之後哪個數字動了,AI 以 Tech 視角改寫成這一頁等級的英文 bullet。四項分數由規則引擎計算,同一份稿重測數字不變。缺的數字一律留黃色 [N] 空格由你自己填,因為面試時要替那些數字辯護的人是你。
免費生成 2 次接著讀
同一位的英文求職信
信只引用履歷上的事實,逐段說明每一句在做什麼,旁邊是量測引擎現算的信件檢查。
資料科學家的英文求職信怎麼寫