CV
Download the PDF or read the full CV below.
Contact Information
| Name | Pragyananda Sahoo |
| Professional Title | Junior Research Fellow (AI/ML) · NLP and Document AI |
| pragyanandasaho@gmail.com | |
| Phone | +91 6398665353 |
| Website | https://pragyananda.github.io |
Professional Summary
Junior Research Fellow at Swami Rama Himalayan University and former AI Engineer at CAIR Lab, DSVV. I build LLM, RAG, and computer-vision systems for document-heavy domains (legal, patents, historical books) and publish on machine-generated text detection, narrative similarity, and vision-based Ayurvedic diagnosis.
Experience
-
2026 - present Doiwala, Uttarakhand, India
Junior Research Fellow (AI/ML)
Swami Rama Himalayan University
- Continuing development of HistoAI, the historical-books ecosystem started at CAIR Lab, DSVV (OCR with image extraction, structured data extraction, knowledge-graph generation, and a RAG chatbot over the digitised collection).
- Conducted an NLP study on child-safe advertisements; paper submitted to the Natural Legal Language Processing (NLLP) Workshop (under review).
- Built web application prototypes and contributed to other research projects at the university.
-
2024 - 2026 Haridwar, India
AI Engineer
CAIR Lab, Institute of AI, Dev Sanskriti Vishwavidyalaya
- Built LLM and RAG products for legal, patent, and historical-document workflows - Supreme Court legal case matching, Innopat patent drafting, and HistoAI.
- Member of the AI-Monitors team for the COLING 2025 GenAI Content Detection task and SemEval-2026 Task 4 (narrative similarity).
- Applied research on generative AI for images, video, and text, transformer architectures, and LLM fine-tuning; trained and deployed models for real-world use and mentored junior developers.
-
2024 - 2024 Haridwar, India
Project Intern (Computer Vision)
CAIR Lab, Institute of AI, Dev Sanskriti Vishwavidyalaya
- Researched handwritten document transformation using document AI transformers and open-source OCR models such as docTR; built a handwritten-text OCR web app with the MERN stack and Flask.
- Built a face-recognition attendance system with Tkinter, OpenCV, and Haar cascades.
- Integrated OpenAI APIs and worked with LangChain and Azure services.
-
2023 - 2023 Haridwar, India
Internship Trainee (Web Development)
Travel Vaidya
- Migrated uttarakhandtrips.com from WordPress to a scalable Django platform with payment integration, MySQL, and a REST API.
-
2022 - 2022 Haridwar, India
Summer Intern (Web Development)
Dev Sanskriti Vishwavidyalaya
- Built internal web applications supporting university operations with PHP and MySQL, including e-commerce features.
Education
-
2022 - 2024 Haridwar, India
Master of Computer Applications (MCA)
Dev Sanskriti Vishwavidyalaya
Computer Applications, Data Science specialization · CGPA 8.20/10
-
2019 - 2022 Haridwar, India
Bachelor of Computer Applications (BCA)
Dev Sanskriti Vishwavidyalaya
Computer Applications, Computer Science · CGPA 7.42/10
Publications
-
2026 AI-Monitors at SemEval-2026 Task 4: A Hybrid Embedding and LLM Ensemble for Narrative Similarity
Proceedings of the 20th International Workshop on Semantic Evaluation (SemEval-2026), ACL, pp. 2139-2148
Determines which of two candidate stories is more narratively similar to an anchor. Structured few-shot prompting and an error-diversity ensemble reached 75% test accuracy, ranking 3rd of 47 systems, close to the 78% human annotator ceiling.
-
2026 Marma Point Detection and Personalized Ayurvedic Diagnosis Using Computer Vision
Indian Knowledge System and Wellbeing: Proceedings of MBCC 2025, Volume 1, Springer Nature, p. 228
First-author paper. A methodology that analyses user images with body type (Vata, Pitta, Kapha), height, symptoms, and medical history to recommend personalised marma points for stress and anxiety therapy. Presented at the 2nd MBCC 2025 conference, IIT Mandi.
-
2025 AI-Monitors at GenAI Detection Task 1: Fast and Scalable Machine Generated Text Detection
Proceedings of the 1st Workshop on GenAI Content Detection (GenAIDetect), COLING 2025, Abu Dhabi, pp. 230-235
Fine-tuned RoBERTa and XLM-R models to detect machine-generated text. Careful choice of training epochs, input size, and class-imbalance handling gave a 5-6% absolute gain in accuracy and F1.
Projects
-
HistoAI - Ecosystem for Historical Books
End-to-end pipeline for digitising historical books, developed at CAIR Lab, DSVV and continued at SRHU - OCR with image extraction for degraded scans, structured data extraction, knowledge-graph generation, and a RAG chatbot over the collection.
-
Innopat - AI Patent Drafting Assistant
Turns claim lists and Invention Disclosure Forms into full patent specifications with RAG and LLMs; refined prompts for consistent terminology, PDF/DOCX parsing, and figure mappings.
-
Supreme Court Legal Case Matching System
Vector-based RAG with LLMs that identifies commonalities, semantic similarities, and legal precedents between newly filed and historical cases.
-
AURA - Tax Automation for Ride-share Drivers (Australia)
Accounting automation for UBER and DIDI drivers - extraction of ride-sharing activity statements, pipelines for CBA, NAB, and ANZ bank data, a BAS automation engine, transaction reconciliation, and Excel-integrated reporting.
-
DVAS
Freelance e-commerce store for eco-friendly areca palm leaf cutlery. Next.js 15, Tailwind CSS, Node.js, MongoDB.
-
UttarakhandTrips
Travel booking site migrated from WordPress to Django with payments, MySQL, and a REST API.
Skills
Achievements
Certificates
- AWS Builders Online Series - Amazon Web Services
- Deploy a Django Web App with AWS Elastic Beanstalk (2023)
- Data Analytics with Specialization in Tableau
- Intro to Programming - Kaggle