Data Scientist - Python, ML, LLMs

AKHILA TATIPALLY

Detroit, MI 947-***-**** ******************@*****.***

linkedin.com/in/akhila-tatipally github.com/akhila-tatipally SUMMARY

Data scientist with experience using Python, SQL, statistical analysis, and machine learning to solve research and business problems. Developed predictive models for customer retention, demand forecasting, and automated narrative assessment. Hands-on experience with feature engineering, model evaluation, large language models, and reproducible experimentation using Scikit-learn, PyTorch, MLflow, and cloud platforms.

EDUCATION

University of Michigan–Dearborn Dearborn, MI

Master of Science in Computer and Information Science Dec 2025 Sreenidhi Institute of Science and Technology Hyderabad, India Bachelor of Technology in Electronics and Computer Engineering Jun 2023 EXPERIENCE

Data Scientist University of Michigan–Dearborn Jan 2026 – Present Dearborn, MI

• Developed AI-powered narrative assessment pipelines that compared automated predictions with expert ratings, achieving ICC values of 0.81–0.91 across selected assessment dimensions.

• Fine-tuned LLaMA, Mistral, and Qwen models using LoRA/PEFT, evaluating model configurations and prediction errors to improve scoring consistency.

• Automated scoring and evaluation workflows, reducing manual expert review effort by approximately 90%.

• Applied structured output schemas and score-range validation to standardize predictions and identify outputs requiring further review.

• Tracked model configurations, evaluation metrics, and artifacts using MLflow, supporting reproducible experimentation and model comparison.

• Integrated model inference with containerized API services and batch workflows to make predictions available to downstream applications.

Graduate Research Assistant University of Michigan–Dearborn Jan 2025 – Dec 2025 Dearborn, MI

• Cleaned and consolidated approximately 1,900 expert-labeled narratives, preparing text, word counts, and assessment labels for analysis.

• Built TF-IDF and regression baselines and evaluated transformer-based text representations to establish reference performance for narrative scoring.

• Conducted exploratory data analysis on score distributions, text characteristics, and prediction discrepancies to investigate model limitations.

• Compared baseline models using RMSE and MAE and documented findings to guide subsequent modeling experiments. AI Engineer Product Manager Accelerator Dec 2023 – Dec 2024 Remote, USA

• Developed NLP workflows for scientific document analysis, combining text extraction with BERT, SciBERT, BioBERT, and Longformer models.

• Prepared labeled examples and structured document features for screening and information extraction, contributing to approximately 68% less manual screening effort.

• Applied PICO-based information extraction to organize study characteristics and support evidence synthesis workflows.

• Supported meta-analysis automation involving effect sizes and heterogeneity, preparing extracted study information for downstream analysis.

• Integrated model outputs through REST APIs and contributed validation checks and documentation for repeatable application workflows. Data Scientist Gear Head Motors Jan 2022 – Nov 2023 Hyderabad, India

• Developed customer churn models using XGBoost and Random Forest, supporting targeted retention initiatives that improved retention by approximately 15%.

• Used Python and SQL to clean, join, and analyze customer, sales, service, and operational datasets.

• Built demand forecasting models using lag, rolling-window, and seasonal features to support operational planning.

• Created ETL workflows processing approximately 1,000–5,000 daily telemetry records for analytics and model preparation.

• Applied hypothesis testing and A/B testing to assess business initiatives and communicate findings to stakeholders. PROJECTS

Multivariate Rainfall Forecasting

• Developed LSTM, GRU, and XGBoost models to predict rainfall across 1-, 3-, and 6-day horizons using historical weather observations.

• Prepared time-series sequences and compared predictions using precision, recall, F1, and accuracy to assess performance on imbalanced rainfall outcomes.

SkinScope AI – Retrieval-Augmented Question Answering

• Combined OCR, ingredient normalization, and hybrid retrieval using vector search and BM25 to retrieve relevant information from product content.

• Implemented an Answer–Critic–Rewrite workflow, reducing observed hallucinations from approximately 18% to 4% in project evaluations.

SKILLS

• Programming and Analysis: Python, SQL, R, Pandas, NumPy, exploratory data analysis, data cleaning

• Statistics: Hypothesis testing, A/B testing, regression, experimental design, error analysis

• Machine Learning: Classification, regression, clustering, XGBoost, Random Forest, Scikit-learn, feature engineering, time-series forecasting

• Deep Learning and NLP: PyTorch, TensorFlow, Hugging Face, BERT, LLaMA, Mistral, Qwen, LoRA/PEFT, LSTM, GRU

• Generative AI: RAG, prompt engineering, structured outputs, LangChain, LangGraph, FAISS, Pinecone, BM25

• Deployment and Tools: MLflow, FastAPI, Docker, Git, CI/CD, AWS SageMaker, S3, Lambda, Vertex AI, Databricks

Contact this candidate