Manasvi Tallapalli Email & Phone Number

Austin, Tx, Us

Austin, Tx, Us

Pleasanton, California, United States

- Collaborated with interdisciplinary teams to develop a RAG-based support assistant, reducing query resolution time by 40%.- Optimized interaction flow using LangChain, enhancing the assistant’s natural language understanding capabilities.- Applied prompt engineering techniques to fine-tune responses, achieving a 30% improvement in query resolution effectiveness, as monitored by accuracy and user satisfaction metrics.- Implemented GPT-4 embeddings with Pinecone and Elasticsearch for rapid information retrieval, boosting response accuracy by 30% and improving customer satisfaction by 25%.- Performed hyperparameter tuning on GPT-4, focusing on learning rate, batch size, and embedding dimensions. Validated performance using cross-validation to ensure consistency across datasets.- Integrated a Streamlit interface for real-time monitoring and interaction- Deployed on Azure Kubernetes Service (AKS) using Docker, supported by A/B Testing for model validation and Jenkins for CI/CD, achieving 99.9% uptime.

Manhattan Beach, California, United States

- Developed a React-based Data Catalog hosted on IIS, integrating Data Lineage, Metadata, Business Glossary, Data Descriptions, and Reference Links, improving data understanding and retrieval efficiency by 60%.- Conducted market research (Gartner, Forrester), to benchmark incorporate industry best practices into the catalog design.- Designed a 3-tier architecture and data models (logical, physical, and conceptual) using Lucidchart ensuring scalability, maintainability, and alignment with existing systems.- Collaborated with data engineers and scientists to implement Data Lineage visualizations using ReactFlow with JSON files from Snowflake SQL scripts (parsed with sqlparse), experimenting with Networkx, Plotly, and D3.js for optimal clarity.- Built the Metadata section, processing and analyzing data from Databricks DBFS (config files) and Snowflake.- Engineered the Data Descriptions section by fine-tuning and evaluating LLMs (OpenAI, Huggingface Transformers, DialoGPT-medium, flan-T5-small, flan-u2-dolly-lora, llama_index). Selected flan-T5-small for superior performance, genrating intelligent summaries and enhancing data accessibility by 30%.- Integrated flan-t5-small into Reference links, connecting users to Confluence, Astronomer, and JIRA.- Created a Figma prototype of the Catalog, enabling stakeholders to interact, provide feedback, and refine the design.- Documented the tool on Confluence for visibility and future reference.- Led weekly project updates, ensuring alignment with engineers, scientists, and senior management.- Engaged in organizational activities (design building, warehouse, retail store tours, discussions with CFO/VPs) to enhance business understanding for catalog design.- Enhanced professional skills through Toastmasters and Snowflake Foundations training, optimizing data utilization and communication.

Hyderabad, Telangana, India

- Automated financial report ingestion (PDF and XBRL files) using AWS S3 for storage and AWS Lambda for workflows. Utilized AWS Textract for OCR and Python (BeautifulSoup4) for parsing XBRL files.- Processed and transformed financial data with Python, storing structured data in Amazon RDS and unstructured data in DynamoDB, reducing query response times by 40%.- Orchestrated workflows with AWS Step Functions for efficient data ingestion, processing, and storage.- Conducted financial analysis using MySQL and scikit-learn to extract key metrics, detect anomalies, and identify red flags improving detection accuracy by 35%.- Evaluated NLP models (BERT, RoBERTa, DistilBERT) for report summarization and Long Context Question Answering. Used ROUGE metrics to validate the quality of generated summaries, enhancing summarization accuracy by 20%.- Integrated BERT with Elasticsearch for efficient search and deployed a Flask API for real-time query processing.- Collaborated with financial analysts to seamlessly integrate the data pipeline, enhancing operational efficiency.

Hyderabad, Telangana, India

• Designed scalable ETL pipelines using AWS Glue, PySpark, and Amazon S3 to process large volumes of client data.• Improved accuracy and processing time by 30%, integrating sources from MySQL, DynamoDB, and Redshift.• Implemented interactive Tableau dashboards for reporting, leading to a 20% improvement in operational efficiency.• Conducted A/B testing and statistical techniques like hypothesis testing and ANOVA on marketing campaigns and pricing strategies to identify KPIs using data from MySQL and Amazon DynamoDB, increasing ROI by 15%.• Developed XGBoost and LSTM models to predict churn and sales, on AWS SageMaker, improving retention by 25%.

Master Of Science - Ms, Data Analytics Engineering

Btech - Bachelor Of Technology, Computer Science

High School, 11Th And 12Th Grades

10Th, Icse

Activities and Societies: Member of gramaseva committee in 8th grade, Member of Activity committee in 9th grade, Member of Art and Culture.