Shubham Chakraborty

Arizona State University|Tempe, AZ

• •Built comprehensive RAG system combining vector databases with fine-tuned LLMs, achieving 78% improvement in answer accuracy and reducing hallucinations by 65% for academic information retrieval tasks.
• •Conducted extensive prompt engineering experiments across cutting-edge foundation models (GPT-4, Claude, Llama-2), improving response quality by 37% through systematic optimization of query patterns and context windows.

Nexus Info|Coimbatore, IN

• •Developed scalable BERT-based NLU system achieving 95% accuracy across 12+ customer intent categories and reducing response times from 15 to 5 minutes for 1000+ daily customer service interactions.
• •Optimized end-to-end ML pipeline on AWS infrastructure, reducing model serving costs by 20% and processing latency by 40% while maintaining prediction quality through efficient batch processing and caching.
• •Implemented advanced data augmentation techniques, increasing training samples by 300% and improving model robustness to unseen queries by 25% through comprehensive distribution testing and validation.

Arizona State University|Tempe, AZ

• •Architected a responsive UI with WCAG 2.1 compliance for the LTH engineering website, resulting in 30% faster page load times and 95% reduction in accessibility issues across 100+ dynamic pages.
• •Implemented custom CMS plugins and RESTful API integrations with the Canvas LMS, reducing faculty course publishing time by 45% while enabling seamless collaboration for 50+ instructors across departments.
• •Developed interactive analytics dashboards using React, D3.js, and Redux for real-time user engagement tracking, driving a 20% increase in site traffic and producing actionable insights for content optimization.

Varcons Technologies|Bengaluru, IN

• •Engineered responsive data visualization components using React, TypeScript, and D3.js for real-time dashboards, improving data visibility by 40% and enabling faster decision-making for business stakeholders.
• •Developed scalable RESTful APIs using Node.js, Express, and MongoDB for efficient data retrieval, reducing average page load times by 30% and enhancing overall user experience across web applications.
• •Implemented a robust CI/CD pipeline with GitHub Actions, achieving 90% test coverage through Jest and Cypress, resulting in 75% reduction in QA time and accelerated release cycles.

Emitrix.io|Remote

• •Built Sustain AI using OpenAI's GPT-3.5 API and custom Python scripts to parse environmental data, reducing report generation time by 40% for 8 enterprise clients while securing $100K seed funding.
• •Trained Random Forest and XGBoost models on client energy consumption patterns, identifying peak usage times and suggesting optimizations that reduced consumption by 30%.
• •Developed custom spaCy NER model to extract compliance metrics from documents, automating 50% of manual reporting tasks.
• •Implemented MLOps using Docker containers on AWS ECS with auto-scaling and CloudWatch monitoring for 99% uptime.