Data Scientist Resume Examples — Junior to Principal Level

SARAH CHEN San Francisco, CA | [email protected] | (415) 555-0142 linkedin.com/in/sarahchen-ds | github.com/sarahchen-ml SUMMARY Data Scientist with 1.5 years of experience building production ML models in Python and deploying on AWS SageMaker. Delivered a demand forecasting model processing 850K daily SKU predictions with 94.3% accuracy, reducing overstock waste by $1.8M annually. Proficient across the full ML pipeline from data ingestion (SQL, Spark) through model training (scikit-learn, XGBoost, TensorFlow) to deployment (Docker, SageMaker endpoints). TECHNICAL SKILLS Languages: Python, SQL, R ML/AI: scikit-learn, XGBoost, LightGBM, TensorFlow 2.x, Hugging Face Transformers Data Engineering: pandas, PySpark, Apache Airflow, dbt Cloud & MLOps: AWS SageMaker, S3, Redshift, Docker, MLflow, Git Visualization: Tableau, Matplotlib, Seaborn, Plotly Databases: PostgreSQL, Snowflake, Google BigQuery EXPERIENCE Data Scientist | RetailTech Inc. | San Francisco, CA | Jun 2024 – Present - Built demand forecasting model using LightGBM on 14 months of transaction data (47M rows, 850K daily SKU predictions), achieving MAE of 3.2 units vs. prior heuristic MAE of 8.7 units — reduced overstock waste by $1.8M/year - Engineered 63 features from raw POS and inventory data using PySpark and dbt, improving model AUC from 0.81 to 0.91 on holdout validation set - Deployed model to AWS SageMaker real-time endpoint with 42ms P99 latency, serving predictions to 12 retail locations via REST API - Designed A/B test framework comparing ML-driven vs. rule-based restocking across 4 pilot stores; ML approach showed 23% reduction in stockouts (p < 0.01, n = 34,200 SKU-days) - Automated weekly model retraining pipeline using Apache Airflow, reducing data-to-prediction cycle from 3 days to 4 hours - Tracked all experiments in MLflow with reproducible configs; maintained model registry with 14 versioned models across 3 use cases Data Science Intern | FinanceAI Corp. | New York, NY | May 2023 – Aug 2023 - Developed credit risk classification model using XGBoost on 2.1M loan applications, achieving AUC of 0.87 and Gini coefficient of 0.74 on out-of-time validation - Reduced false positive rate by 18% (from 12.3% to 10.1%) through SHAP-based feature selection, removing 22 multicollinear features without degrading AUC - Built interactive Tableau dashboard displaying model performance metrics across 8 risk segments for the credit review committee (15 stakeholders) - Wrote SQL queries processing 500GB of transaction data in Snowflake with average execution under 12 seconds using optimized partitioning EDUCATION M.S. Data Science | Columbia University | New York, NY | 2024 - Thesis: "Attention-Based Feature Selection for Tabular Data" — presented at Columbia Data Science Society - Relevant Coursework: Statistical Machine Learning, Deep Learning, Bayesian Statistics, Causal Inference B.S. Statistics, Minor in Computer Science | UC Berkeley | Berkeley, CA | 2022 - GPA: 3.78/4.0 | Dean's List 6 semesters CERTIFICATIONS - AWS Certified Machine Learning — Specialty, Amazon Web Services (2024) - TensorFlow Developer Certificate, Google (2023) PROJECTS - Open-source contributor to scikit-learn: merged PR #28431 improving HistGradientBoosting NaN handling (github.com/sarahchen-ml/sklearn-contrib) - Kaggle Competition: Jane Street Market Prediction — Top 4% (Silver Medal, 187th / 4,245 teams)

MARCUS JOHNSON Seattle, WA | [email protected] | (206) 555-0387 linkedin.com/in/marcusjohnson-ml | github.com/mjohnson-ds SUMMARY Senior Data Scientist with 4 years of experience designing and deploying ML systems at scale. Led development of a real-time fraud detection platform processing 2.4M transactions/day with 97.2% precision at 0.003% false positive rate, preventing $31M in annual fraud losses. Expert in end-to-end ML pipelines from feature stores (Databricks, Feast) through model serving (SageMaker, Vertex AI) with strong focus on experimentation platforms and causal inference. TECHNICAL SKILLS Languages: Python, SQL, Scala, R ML/AI: PyTorch, TensorFlow, scikit-learn, XGBoost, LightGBM, Hugging Face, LangChain Data Engineering: Apache Spark, Apache Airflow, dbt, Kafka, Flink Cloud & MLOps: AWS (SageMaker, Redshift, Glue, Lambda), GCP (Vertex AI, BigQuery, Dataflow), MLflow, Weights & Biases, Docker, Kubernetes Feature Stores: Databricks Feature Store, Feast Visualization: Tableau, Looker, Streamlit Databases: Snowflake, PostgreSQL, Redis, DynamoDB EXPERIENCE Senior Data Scientist | PaySecure (Series C Fintech, $2.1B valuation) | Seattle, WA | Mar 2023 – Present - Architected real-time fraud detection system using gradient-boosted ensemble (XGBoost + LightGBM) with streaming feature computation via Kafka and Flink, processing 2.4M transactions/day at 97.2% precision and 0.003% false positive rate - Prevented $31M in annual fraud losses while maintaining sub-15ms inference latency per transaction on SageMaker multi-model endpoints - Built and maintained feature store with 340+ features (Databricks Feature Store) including 48 real-time features computed from 90-day sliding windows of user behavior - Designed experimentation platform enabling 12 concurrent A/B tests; proved that dynamic risk thresholds increased fraud catch rate by 14.7% (95% CI: [12.1%, 17.3%]) vs. static rules - Developed customer lifetime value model using survival analysis (Cox proportional hazards) on 1.8M user histories, identifying $4.7M in at-risk revenue for proactive retention campaigns - Mentored 2 junior data scientists through code reviews, pair programming sessions, and weekly ML paper discussions Data Scientist | DataDriven Health (Digital Health Startup) | San Francisco, CA | Jul 2021 – Feb 2023 - Built patient readmission prediction model using PyTorch tabular transformer on 890K patient encounters (230+ features from EHR data), achieving AUC of 0.84 — a 12-point improvement over the logistic regression baseline (AUC 0.72) - Reduced 30-day readmission rate by 9.3% across 3 partner hospitals by integrating model predictions into clinical workflow dashboards, translating to $2.8M annual savings in CMS penalties - Engineered NLP pipeline using BioBERT to extract structured diagnoses from 1.4M unstructured clinical notes, achieving F1 of 0.91 on 23 ICD-10 code categories - Optimized Spark data pipeline processing 4.2 TB of claims data from raw ingestion to feature-ready tables, reducing nightly batch runtime from 6.5 hours to 47 minutes through partition pruning and predicate pushdown - Created model monitoring dashboard in Streamlit tracking PSI (Population Stability Index) drift across 12 feature distributions, triggering automated retraining when PSI exceeded 0.2 threshold Data Analyst → Data Scientist | TechCommerce Inc. | Austin, TX | Jun 2020 – Jun 2021 - Promoted from Data Analyst to Data Scientist within 8 months based on delivery of product recommendation engine that increased average order value by 11.4% ($8.20 per order) across 340K monthly active users - Built collaborative filtering model (matrix factorization via implicit library) trained on 28M click-purchase events, serving personalized recommendations via Redis cache with 6ms average response time - Conducted 14 A/B tests on recommendation placement and algorithm variants, generating $3.1M incremental annual revenue from the winning configurations - Migrated reporting infrastructure from Excel-based manual processes to automated Airflow + dbt pipeline delivering 23 daily Tableau dashboards to 45 stakeholders EDUCATION M.S. Computer Science (Machine Learning Specialization) | Georgia Institute of Technology | 2020 - Published: "Efficient Feature Selection via Mutual Information Networks" — ICML Workshop on AutoML, 2020 B.S. Mathematics | University of Texas at Austin | 2018 CERTIFICATIONS - Google Professional Machine Learning Engineer, Google Cloud (2023) - AWS Certified Machine Learning — Specialty, Amazon Web Services (2022) - Microsoft Certified: Azure Data Scientist Associate, Microsoft (2023) PUBLICATIONS & SPEAKING - "Streaming Feature Engineering for Real-Time Fraud Detection" — KDD Applied Data Science Track, 2024 - Speaker, MLconf Seattle 2023: "From Batch to Real-Time: Migrating ML Pipelines Without Downtime"

PRIYA RAMANATHAN, Ph.D. New York, NY | [email protected] | (212) 555-0291 linkedin.com/in/priya-ramanathan-phd | github.com/priya-ml | scholar.google.com/priya-ramanathan SUMMARY Principal Data Scientist and ML Engineering Leader with 8 years of experience building ML platforms that drive $120M+ in measurable business value. Led a team of 11 data scientists and ML engineers to design, deploy, and monitor 40+ production models across pricing, personalization, and fraud detection. Deep expertise in ML system design (feature platforms, model serving, experimentation infrastructure), LLM applications (RAG, fine-tuning, evaluation), and translating complex technical capabilities into executive-level business strategy. TECHNICAL SKILLS Languages: Python, SQL, Scala, Julia, C++ ML/AI: PyTorch, TensorFlow, JAX, scikit-learn, XGBoost, Hugging Face Transformers, LangChain, vLLM LLM & GenAI: GPT-4, Claude, Llama 2/3, RAG architectures, RLHF, prompt engineering, LLM evaluation frameworks Data Engineering: Apache Spark, Apache Kafka, Apache Flink, dbt, Airflow, Dagster, Delta Lake Cloud & MLOps: AWS (SageMaker, Bedrock, EMR, Glue, Redshift), GCP (Vertex AI, BigQuery, Dataflow), Databricks, MLflow, Weights & Biases, Kubeflow, Docker, Kubernetes Databases: Snowflake, PostgreSQL, Redis, Pinecone, Weaviate Visualization: Tableau, Looker, Streamlit, Grafana EXPERIENCE Principal Data Scientist & ML Platform Lead | MegaRetail Corp. (Fortune 100, $48B revenue) | New York, NY | Jan 2022 – Present - Lead team of 11 data scientists and 4 ML engineers responsible for all ML-driven pricing, personalization, and supply chain optimization — models collectively influence $4.8B in annual GMV - Designed and deployed dynamic pricing engine using contextual bandits (Thompson Sampling) across 2.1M SKUs, driving $47M incremental annual margin (3.2% lift) with <200ms latency per price computation on Kubernetes-hosted model serving - Architected centralized feature platform on Databricks with Delta Lake, serving 1,200+ features to 40+ production models with 99.95% uptime SLA — reduced new model onboarding time from 6 weeks to 8 days - Built enterprise RAG system for internal knowledge retrieval using Llama 3 70B fine-tuned on 2.3M company documents, deployed via vLLM on 8x A100 GPU cluster; achieved 91% answer accuracy (human-evaluated, n=2,400) reducing analyst research time by 62% - Established ML governance framework including model cards, bias audits (demographic parity checks across 6 protected classes), and automated monitoring for 40+ models — framework adopted company-wide across 3 business units - Presented quarterly ML impact reviews to C-suite (CEO, CFO, CTO), translating model performance metrics into $120M cumulative P&L impact narrative that secured $8.5M annual ML infrastructure budget Senior Data Scientist | AlgoTrading Partners (Quantitative Hedge Fund, $3.2B AUM) | New York, NY | Mar 2019 – Dec 2021 - Developed alternative data signals using NLP (BERT fine-tuned on 4.7M SEC filings and earnings call transcripts), generating alpha signal with Information Coefficient of 0.08 — top-decile among 34 internal signal sources - Built portfolio optimization model combining 12 ML-derived signals with classical mean-variance framework, contributing to $180M in annual trading revenue for the systematic equities desk - Engineered low-latency feature pipeline processing 14 TB/day of market tick data through Kafka and Flink, delivering 230+ features to model serving layer with end-to-end latency under 850ms - Reduced model training time by 73% (from 11 hours to 3 hours) by migrating from single-GPU PyTorch to distributed training on 4x A100 nodes using PyTorch DistributedDataParallel - Designed backtesting framework with walk-forward validation across 8 years of market data, incorporating realistic transaction costs and slippage — eliminated 3 spurious signals that showed +2.1% annualized alpha in naive backtest but -0.4% in realistic simulation Data Scientist II | InsuranceAI (Insurtech, acquired by Prudential for $890M) | Chicago, IL | Aug 2017 – Feb 2019 - Built auto insurance pricing model using gradient-boosted trees (XGBoost) on 6.4M policy records with 180+ features including telematics data, achieving 15% improvement in loss ratio prediction (Gini from 0.31 to 0.46) - Developed claims severity model processing 50K monthly claims with 92.4% accuracy in severity bucket classification, enabling $12M reduction in reserve estimation error annually - Created churn prediction system identifying at-risk policyholders 90 days before renewal, achieving precision of 0.78 at 30% recall — enabled targeted retention campaigns that recovered $6.3M in annual premium - Automated actuarial data pipeline in Airflow processing 3.8 TB of claims, policy, and telematics data nightly; reduced manual actuarial prep from 12 hours/week to zero Data Scientist | StartupML (Y Combinator W2016) | San Francisco, CA | Jun 2016 – Jul 2017 - First data hire at Series A startup; built ML pipeline from scratch using scikit-learn and PostgreSQL, delivering product recommendation engine that increased user engagement by 34% (DAU/MAU ratio from 0.21 to 0.28) - Implemented real-time user segmentation processing 2M events/day via Kafka Streams, enabling personalized push notifications that improved 7-day retention by 8.7 percentage points EDUCATION Ph.D. Statistics (Bayesian Nonparametrics) | Stanford University | 2016 - Dissertation: "Scalable Gaussian Process Methods for Large-Scale Spatial-Temporal Prediction" - Published 4 papers in ICML, NeurIPS, and JMLR B.S. Mathematics & Computer Science (Summa Cum Laude) | MIT | 2011 CERTIFICATIONS - Google Professional Machine Learning Engineer, Google Cloud (2023) - AWS Certified Machine Learning — Specialty, Amazon Web Services (2022) - IBM Data Science Professional Certificate, IBM (2021) PUBLICATIONS (Selected) - "Contextual Bandits for Dynamic Retail Pricing at Scale" — KDD 2024 (Oral Presentation) - "Scalable Feature Platforms: Lessons from Serving 1,200+ Features in Production" — MLSys 2023 - "Bias Auditing in Production ML Systems: A Practitioner's Guide" — FAccT 2023 - 12 total peer-reviewed publications | 1,400+ citations (Google Scholar) BOARD & ADVISORY - Technical Advisory Board, Women in Data Science (WiDS) Conference (2022 – Present) - ML Advisory Council, NYU Center for Data Science (2023 – Present)