Sushma Niveni Pindiga

E

AI Specialist (LLM Distillation)]

EHEALTHLINE INC

Mar 2025 - Current (1 year 4 months)

• Designed and implemented a knowledge distillation pipeline transferring step-by-step reasoning from a DeepSeek LLM to a compact BERT model using patient health data. • Generated over 20,000 rationales from the teacher model via Lang Chain, embedding interpretability into the training process of the student model. • Engineered a multi-task learning framework combining classification loss and rationale alignment loss, improving model interpretability by 40% while maintaining prediction accuracy. • Preprocessed clinical features including MMSE, ADL, memory complaints, and cardiovascular indicators, ensuring high data consistency and model readiness. • Achieved a student model AUC of 94.2% and accuracy of 89.6%, closely matching the teac

R

AI (Quantitative Analyst)

REFLECTION

Sep 2024 - Mar 2025 (7 months)

• Applied and validated logistic & linear regression models for fair lending, identifying 6 documentation findings. • Created Tableau dashboards and quarterly reports for 6 quarters to track portfolio performance & sensitivity analysis. • Implemented XGBoost regression model with hyperparameter tuning, achieving RMSE of 0.25 in Python. • Prepared & transformed 35 macroeconomic variables to model Probability of Default, collaborating with cross functional teams. • Supporting a Compliance analysis focused on compliance reporting. • Handling data quality checks, intake tracking, and dashboarding using Tableau and Power BI.

Data Scientist

Masonite

Sep 2023 - Sep 2024 (1 year 1 month)

• Utilized SQL and Python for predictive analysis on 5M POS transactions, optimizing sales strategies and inventory. • Developed a Generative AI solution with Snowflake, Azure AI, Lang Chain (Python), and Streamlit, replacing Power BI dashboards. • Boosted query speed by 150% and increased user interaction by 50%, delivering real-time insights. • Conducted KPI analysis with Excel, reducing data processing time by 25% and error rates by 10%. • Engineered predictive analytics for trend identification, generating 5,000+ daily responses. • Led prompt engineering, model tuning, and UI design, achieving User Satisfaction Score of 7/10.

Software Engineer

DuPont

Jan 2022 - Jul 2023 (1 year 7 months)

• Managed DuPont's portfolio of websites; updates via Adobe Experience Manager (AEM), achieving less than 24-hour content update turnaround, maintaining 99.9% uptime, enhancing user experience. • Improved web standards and accessibility, resulting in a 15% increase in user engagement. • Led team adoption of new AEM features, reducing operational overhead by 20%. • Participated in SDLC, creating 600+ user stories with fewer than 5 User Acceptance Testing (UAT) defects.

Data Scientist

DevTown

Jun 2021 - Jan 2022 (8 months)

• Conducted extensive research on time series analysis and natural language processing (NLP), gathering literature from various sources. • Applied and evaluated time series forecasting models, including ARIMA and Facebook Prophet, on available datasets to predict stock index movements. • Published Time-Series Forecasting: Using Arima Facebooks Prophet Model (ijraset.com) • Authored a Study and Analysis on Machine Learning and Artificial Intelligence for Smart Agriculture IJCRT2211590.pdf

Sushma Niveni Pindiga

About

Detail

Timeline

Résumé