S
Shruti Devlekar
About
Detail
Buffalo, New York, United States
Résumé
Jobs
verified_user
0% verified
Data ScientistDataminrJul 2025 - Current (1 year 4 months)• Built real-time event detection workflows using PySpark and AWS SageMaker, improving alert relevance by 18% across public news and social content streams globally.
• Refined NLP classification models using PyTorch and TensorFlow, improving entity extraction consistency by 15% and supporting faster intelligence assessment across emerging risk events.
• Performed Feature Engineering on streaming datasets through ETL Pipelines, accelerating event ingestion by 24% while improving downstream model readiness and processing efficiency.
• Assessed Generative AI summaries against analyst-reviewed outputs, increasing content relevance by 13% and establishing quality benchmarks for intelligence monitoring workflows.
• Maintained- Data ScientistAtlassianJun 2023 - Jul 2024 (1 year 2 months)• Analyzed Jira and Confluence usage data using Python and SQL, uncovering adoption trends that improved feature prioritization accuracy by 14% across products.
• Conducted Exploratory Data Analysis (EDA) across customer interaction datasets, identifying 17% variation in engagement behavior among enterprise and mid-market user segments.
• Developed Predictive Modeling solutions estimating feature adoption likelihood, improving targeting effectiveness by 12% and supporting roadmap planning decisions for product teams.
• Processed behavioral datasets through Apache Spark, reducing feature preparation time by 21% while supporting recurring analytics and experimentation initiatives across platforms.
• Created Tableau dashboa - Data Analyst (Intern)CleverTapJan 2023 - May 2023 (5 months)• Examined campaign performance data using SQL and Power BI, identifying audience behavior trends that improved targeted engagement effectiveness by 12% across applications.
• Integrated customer interaction records from messaging, push notification, and in-app channels, improving reporting completeness by 19% across marketing analytics datasets.
• Utilized Python, Pandas, and NumPy for Data Wrangling, reducing segmentation preparation effort by 15% while supporting campaign performance analysis initiatives.
• Validated Data Quality and attribution reporting outputs, reducing dashboard discrepancies by 16% and improving confidence in campaign measurement across stakeholders.
Education
verified_user
0% verified
- IIntro to Deep LearningApr 2026
- MMicrosoft Power BIApr 2026
- EExploratory Data Analysis for Machine LearningMar 2026
- TTime SeriesMar 2026
- AAgentic AI with LangChain and LangGraphMar 2026
- FFeature EngineeringMar 2026
- Master of Science, Data ScienceUniversity at BuffaloAug 2024 - Dec 2025 (1 year 5 months)
- VBachelor of Technology, Artificial Intelligence and Data ScienceVivekanand Education Society's Institute of TechnologySep 2020 - May 2024 (3 years 9 months)
Projects (professional or personal)
verified_user
0% verified
- Loan Default PredictionPersonal ProjectJun 2026• Engineered a loan default prediction platform processing 255,000 records, achieving 0.80 ROC-AUC while reducing false positive rates from 24.9% to 10.1%.
• Developed FastAPI prediction services returning probabilities, thresholds, and top-ten SHAP explanations, improving model interpretability and decision-making for risk assessments.
• Containerized application workloads using multi-stage Docker builds and automated GitHub Actions validation pipelines, ensuring consistent deployments and reliable release quality.
• Built an interactive Streamlit dashboard featuring threshold optimization and counterfactual analysis, enabling users to evaluate prediction sensitivity across lending scenarios. - Hospital Readmission MLOps PlatformPersonal ProjectMay 2026• Developed an end-to-end readmission prediction pipeline on 101,766 patient encounters using Optuna-tuned XGBoost, achieving 0.6776 ROC-AUC and 0.2344 PR-AUC.
• Built an MLflow-based model registry with automated promotion gates, enabling reproducible experimentation and preventing performance regressions during retraining workflows.
• Containerized FastAPI inference services using multi-stage Docker builds with HEALTHCHECK monitoring, reducing deployment overhead while improving production reliability and portability.
• Implemented Evidently drift monitoring across 50 features and SHAP explainability outputs, supporting model transparency, data quality assessment, and operational governance. - Telecom Customer Churn IntelligencePersonal ProjectApr 2026• Built a churn prediction system for 7,043 telecom customers using calibrated HistGradientBoosting models, achieving 0.8448 ROC-AUC and 77% minority-class recall.
• Engineered seven predictive features and optimized classification thresholds through StratifiedKFold validation, improving churn detection accuracy across customer segments.
• Developed eight analytical SQL queries leveraging CTEs, window functions, and cohort analysis to quantify revenue risk and retention opportunities.
• Delivered interactive Streamlit and Tableau dashboards identifying $2.86M monthly revenue exposure and modeling approximately $280K net retention ROI. - PPersonalMay 2023 - Jun 2023 (2 months)Recurrent Neural Network●
Publications
verified_user
0% verified
- EEnhancing Dialogue Systems with Adaptive Decision Boundaries and Multi-task Learning for Open Intent RecognitionJan 2025
- AAnomaly Detection in Time Series using Unsupervised Machine Learning ApproachJan 2025
- PPredicting and classifying water quality, treatment, and usage: a comprehensive reviewJan 2025