Kaushal Hebbar Email & Phone Number

Bengaluru, KA, IN

Bengaluru, Karnataka, India

• Conducted prompt engineering experiments to enhance understanding and capabilities of large language models (LLMs). Fine-tuned several open-source LLMs such as Mistral, Flan, and Gemma family models, gaining extensive.
• Created a Modular Visualization Pipeline for validating and visualizing data from documents, incorporating Human In The Loop (HITL) feedback to audit and correct data extractions.
• Developed a real-time… Show more
• Developed a real-time application for extracting medical data from Pillboxes, achieving end-to-end processing latency of ~1 second with ~90% accuracy on 12 key entities using prompt sharding techniques.
• Led and mentored a team of interns, overseeing task assignments and providing guidance to ensure successful project execution. Show less

Bangalore Urban, Karnataka, India

• Spearheaded the development and optimization of a new pipeline for extracting over 400 key-value pairs from complex attachments, cutting processing time from 5 days to 5 minutes using NLP and Vision techniques.
• Engineered a robust data extraction system for medical insurance ID cards, accommodating variations across 30+ carriers with sub-5 second processing times.
• Achieved 90% accuracy and 70% faster processing for FNOL submissions by integrating advanced models for entity… Show more
• Achieved 90% accuracy and 70% faster processing for FNOL submissions by integrating advanced models for entity extraction and data normalization from documents and text.
• Designed a search engine to extract and index information from 650+ Terms & Conditions PDFs, overcoming challenges posed by low-quality OCR scans and multi-columned document formats.
• Created a pipeline for classifying a large volume of emails with attachments, employing advanced text extraction and classification methods. Show less

Bengaluru, Karnataka, India

• Developed a pipeline using an ensemble of deep learning models to extract machine-printed and cursive handwritten text in a single shot, leveraging multiple OCR engines.
• Fine-tuned multimodal models like LayoutLM for accurate recognition of key-value pairs from healthcare and insurance documents.
• Explored and implemented open-source deep learning models for detecting tables and extracting structured data from documents.

Bengaluru, Karnataka, India

Bachelor Of Technology, Electronics And Communications Engineering, 8.99

Minor Degree, Computer Science And Engineering

A Minor B. Tech Degree in CSE focusing on Data Structures and Algorithms, Operating Systems and DBMS.

Pre-University, 98.3%

Karnataka State Board, 10Th, 98.4%