Nithil Eshwar Mani

Master of Science (Data Science) @ University at Buffalo
Graduated : January 2025
Coursework: Statistical Learning, Data Mining, Data Models Query language, Predictive Analysis, Analysis of Algorithms.

Bachelor of Engineering (Computer Science and Engineering) @ College of Engineering, Guindy
Graduated : May 2023
Coursework: Machine Learning, Data Structures, NLP, Big Data Analytics, Database Management Systems.

Software Engineer (Remote)

• Built full-stack applications with Python and Node.js backends and React.js, TypeScript, and CSS frontends to automate paperwork and clinical workflows for healthcare providers.
• Optimized a Text-to-SQL agent to provide SQL results for natural language queries about patients' health records and clinicians' information. Used Vanna AI to train the model and wrote 70+ SQL query prompts. Improved performance on complex queries by 60%.
• Developed an AI-powered call agent to automate patient intake (appointment scheduling, verifying insurance information). Used FastAPI for building APIs and ElevenLabs for conversation handling, storage and retrieval.
• Built dashboards with React, Node.js and Typescript to monitor the apps. Wrote terraform scripts to deploy applications in AWS Amplify and AWS App Runner. Optimized backend code and improved performance by 20%.
• Utilized other AWS services like CloudWatch and Cognito to monitor applications performance and manage user authentication.
• Implemented CI/CD pipelines for automated deployment and infrastructure management and wrote unit and integration testing scripts ensuring reliable performance.

Software Engineer Intern (Remote)

• Built and managed graph databases using Neo4j and MilvusDB to support real-world, large-scale applications.
• Used Neo4j ETL tools generate metadata mappings and export data from PostgreSQL databases.
• Integrated AI models for intelligent relationship discovery and data-driven automation using GraphRAGs.
• Incorporated Serverless applications with simplified development and reduced costs by 50%.

Research Assistant

• Contributed to the development and testing of UnionLabs, a shared online platform to share data, code, software and hardware resources for research in next-generation networks and wireless Internet of Things.
• Established AWS EC2-local server connectivity with API endpoints.
• Conducted thorough testing of back-end, database, and APIs to ensure robustness and reliability of the platform.
• The cloud-based remote-access solution enabled ease and efficiency while reducing lab costs by 60%.
• Improved user accessibility and responsiveness by 70% through optimized React.js front-end code.

• Tech stack: Python, Pandas, Matplotlib, Seaborn, Plotly, Typescript, React.js
• Built a full-stack web application that performs intelligent data analysis, visualization and provides AI-powered insights on CSV files
• Users can upload CSV files and receive comprehensive visualizations including histograms, pie charts, scatter plots, box plots, and heatmaps, along with detailed statistical summaries and column-level insights.
• Fine-tuned the gpt-oss-20b model to automatically provide BI insights, identify important features, determine data types, and recommend univariate and bivariate analyses.

• Tech stack: Python, TensorFlow, SageMaker AI
• Classified over 30000 HEP-TH papers dating from 1991-2004 as influential and non-influential.
• Used AWS SageMaker Blazing Text for classification and Sci-BERT word embedding for tokenising.
• Performed hyperparameter optimization using Amazon SageMaker AI Automatic Model Tuning.
• Obtained 75% accuracy after cross validation.

• Tech stack: Langchain, Cypher, Streamlit
• Built a Q+A agent for Research papers using GraphRAGs, OpenAI’s GPT 4o model and Cypher.
• Designed prompts to create a knowledge graph data source and wrote data retrieval functions using cypher.
• Deployed the application using Streamlit. The GraphRAG allowed for more comprehensive and detailed responses.

• Tech stack: Python, Pytorch, SageMaker AI
• Optimized the Inception-Resnet V1 model to detect Choroidal Neovascularization in over 10000 Retinal OCT images.
• Obtained better accuracy (97.10%) compared to the Amazon SageMaker Image Classification Model (95.43%).
• Trained the model to overcome Elastic-Net attack. Improved the accuracy on adversarial examples from 7.5% to 51.39%.

• AWS Certified Cloud Practitioner
• AWS Certified AI Practitioner
• AWS Certified Data Engineer - Associate
• AWS Certified Machine Learning Engineer - Associate

Worked in a customer-facing environment as a nonprofit representative. I contacted University at Buffalo alumni, helped them by updating their academic and personal information and offered them opportunities to contribute back to the university. I reached out to over 500 alumni, built strong relationship and was able to raise almost 4000$ in 5 months..