Piyush Hinduja

AI Engineer (formerly Data Analyst) at University of Utah

Built language models to analyze Earnings Conference Calls and 10-K, 10-Q, and 8-K filings; implemented advanced NLP analysis using FinBERT, PyTorch, and Transformers for financial document processing; architected an ETL pipeline to extract, parse, and store financial documents from SEC.gov and EDGAR; developed robust prompt engineering solutions to integrate LLMs & GenAIs for automated financial text analysis; improved prediction accuracy from 66% to 94.13%.

AI Engineer at Giant Analytics

Fine-tuned LLaMA 3 model on Indian court data (AWS S3) for domain-specific legal reasoning using AWS SageMaker; integrated the model into a RAG-based SaaS platform that retrieves and cites relevant cases during chat; projected to boost answer accuracy from 56% to 93%, significantly outperforming generic OpenAI models.

Software Developer at Squeeze Media LLC

Built an AI-powered warm transfer system for inbound/outbound Twilio calls using FastAPI, Deepgram SST, and WebSocket streaming; Increased company sales by 28% by ensuring more customers completed their calls without drop-offs or repeated queries.

Full Stack Developer Intern at Exposys Data Labs

Spearheaded a tourism project, integrating key features to enhance user experience and deliver valuable information about 50+ Indian travel destinations; utilized MERN stack for an efficient frontend and backend integration, creating a responsive, user-friendly interface and enabling content management.

AI Engineer (formerly Data Analyst) at University of Utah

Developed NLP-based analysis of Earnings Conference Calls and SEC filings to identify correlations between executive statements and future performance. Implemented FinBERT-based encoder models in PyTorch/Transformers for financial document processing, boosting prediction accuracy from 66% to 94.13%. Built an ETL pipeline to extract, parse, and store documents from SEC.gov and EDGAR, optimizing multi-core/multi-thread processing and reducing per-filing processing time by over 90%. Created robust prompt engineering workflows to integrate LLMs and GenAI (OpenAI, Anthropic, Gemini) for automated financial text analysis and benchmarking against traditional algorithms.

AI Engineer at Giant Analytics

Fine-tuned the LLaMA 3 model on Indian court data for domain-specific legal reasoning; integrated into a RAG-based SaaS platform that retrieves and cites relevant cases during chat, projecting an increase in answer accuracy from 56% to 93% and outperforming generic OpenAI models.

Software Developer at Squeeze Media LLC

Built an AI-powered warm transfer system for inbound/outbound Twilio calls using FastAPI, Deepgram SST, and WebSocket streaming, improving customer call completion rates and reducing drop-offs; contributed to a 28% increase in sales.