Meet Vaghasia

At RootstockLabs, I designed and deployed RAG pipelines that integrated multiple LLM APIs, including GPT, Claude, and LLama, with an Amazon Redshift data warehouse. The system enabled dynamic SQL query generation and delivered instant, formatted insights for blockchain analytics and business reporting.

My role involved architecting a scalable data pipeline that connected diverse data sources through APIs, webhooks, and RPCs. I implemented containerized workflows with Docker and automated validation scripts to ensure data quality and consistency across ETL processes. Additionally, I developed monitoring systems for outlier detection, which improved accuracy and trust in the generated insights.

Through this project, I combined expertise in data engineering, AI integration, and analytics visualization to build a high-impact system that streamlined operations and enabled RootstockLabs to leverage AI for real-time data intelligence in blockchain ecosystems.

At Turing, I worked on optimizing the Gemini LLM for a Google project, focusing on benchmarking, fine-tuning, and improving model performance. I analyzed Gemini’s back-end behavior and compared it with GPT across multiple scenarios to highlight gaps and strengths.

I designed workflows to extract, clean, and transform diverse datasets—CSV, Excel, PDFs, HTML, and text—ensuring high-quality inputs for large-scale training and evaluation. Using advanced transformation and predictive modeling, I improved the consistency and interpretability of outputs.

To enhance alignment, I applied RLHF tools to refine Gemini’s responses and implemented Python-based manual corrections to fix edge cases, which reduced error rates and boosted accuracy. I also built visualizations of performance trends that helped stakeholders track progress and guide iterative improvements.

This project delivered measurable gains in response quality, reliability, and contextual relevance, strengthening Gemini’s competitive position against peer LLMs like GPT.

MV.tech Solutions is my company website for presenting engineering services across data platforms, web applications, business apps, automation, analytics, cloud architecture, and AI systems. I designed and built the site to explain the service areas, expertise, delivery process, and production-minded approach in a clear, well-structured way for potential clients. My work included content organization, responsive UI, service messaging, and a polished web experience that makes the company's technical offering easy to understand.