Nisarg Nargund
Founder of OpenRAG Innovations and published researcher (IJCNN 2026, IEEE, ACM, Springer). I build production AI infrastructure - from efficient LLM inference and quantization to RAG systems that work reliably in the real world.
Professional Experience
Researcher — International Institute of Information Technology (IIIT), Hyderabad, India
May 2025 – July 2025
- Implemented BitNet (1-bit LLM) from scratch; evaluated performance under resource-constrained inference on T4 and H100 GPUs.
- Conducted an in-depth architectural analysis of DeepSeek-V3, studying its MoE design, expert routing, and training efficiency innovations.
- Explored edge computing deployment strategies for low-precision LLMs, informing subsequent quantization research (TernaryLM, IJCNN 2026).
Generative AI Engineer — Dubai-based Tech Company (Remote)
Aug 2025 – Sept 2025
- Built scalable GenAI applications including custom RAG chatbots and LLM pipelines for global clients.
- Optimized prompt engineering and model outputs to reduce hallucinations and improve multilingual support.
- Contributed to production-grade solutions across education and enterprise domains.
From Research to Reality: Entrepreneurial Ventures
OpenRAG Innovations Pvt. Ltd.
Founder & Director
OpenRAG is the AI trust layer that verifies outputs before delivery — making LLMs decision-safe for regulated and high-stakes use cases.
Everyone is deploying AI today — but almost no one is measuring how correct those responses are, or how defensible they are when reputational risk is on the line. OpenRAG solves that. Our system lets you query your own data (Excel, DOCX, PDF) through a persona selection layer built for Fintech and Edtech — so you get role-calibrated, context-aware responses instead of the generic outputs every LLM serves by default.
Before any response reaches the user, our system verifies and scores the output pre-delivery — a step almost all companies skip entirely or perform after the fact. DocDynamo isn't a Q&A tool. It's a decision support system that delivers only trusted, persona-backed outputs — giving enterprises a defensible, auditable AI layer they can actually stand behind.
DocDynamo
Multi-agentic backend with a proprietary hallucination removal and verification layer that sits between human and AI. 2,600+ beta users. 2 active B2B contracts.
Conference Partnerships
Official AI partner — ICMEET 2025 (London) and ICDECT 2025 (Bhubaneswar).
CIN: U58201OD2025PTC050816 · Incorporated: 22 Sept 2025 · support@openrag.in
Research Papers
TernaryLM: Memory-Efficient Language Modeling via Native 1.5-Bit Quantization with Adaptive Layer-wise Scaling
Status: Accepted · Camera Ready
International Joint Conference on Neural Networks, Maastricht, Netherlands, 2026
Demonstrates native 1.5-bit quantization with per-layer precision scaling — enabling competitive LLM performance at extreme memory compression for edge deployment.
Optimization of Latent-Space Compression using Game-Theoretic Techniques for Transformer-Based Vector Search
Status: Accepted
A Transformer-Based Model for Enhanced Tokenization and Generation in Hindi Natural Language Processing
Status: Presented (Pending Publication)
Neural Orchestration for Multi-Agent Systems: A Deep Learning Framework for Optimal Agent Selection
Status: Presented (Pending Publication)
Model Context Protocol (MCP): A Lightweight, Modular Framework for Tool-Augmented LLM Agents
Status: Accepted (Camera Ready)
From Chalkboards to Chatbots: Exploring Teacher and Student Readiness for Agentic AI in Indian Schools
Status: Presented (Pending Publication)
Conversational Text Extraction with Large Language Models Using Retrieval-Augmented Systems
Status: Published
Innovative Fusion of LSTM and Bi-GRU Networks for Enhanced Hate Speech Detection in Social Media
Status: Published
Deep Learning in Industry 4.0: Transforming Manufacturing through Data-Driven Innovation
Status: Published
Advancements in Computer Vision and Machine Learning for Food Quality Evaluation: A Comprehensive Review for the Food Industry
Status: Published
Articles
2024
Beyond Reality: Virtual Worlds Shaping the Future of Food Science
Food Infotech Magazine — January 2024
Read Article →Advancements in Food Safety and Quality Evaluation Using Computer Vision and Machine Learning
Food, Marketing and Technology Magazine — March 2024
Read Article →Implementing the Technological Intelligence in Agricultural Produce
Food, Marketing and Technology Magazine — June 2024
Read Article →Edible and Biodegradable Packaging Innovations
Food, Marketing and Technology Magazine — December 2024
Read Article →2023
Two Layer Classifiers of MVS for Effective Grading, Safety & Quality Evaluation in Food Industry
Food Infotech Magazine — April 2023
Read Article →Crunching the number of safer foods: how big data is transforming food safety
Food Infotech Magazine — May 2023
Read Article →Cracking the Code! Unveiling the Hidden World of Food Safety with MicroWaves & ML
Food Infotech Magazine — June 2023
Read Article →Revolutionizing Agri-Food Supply Chain: Harnessing the Power of IoT
Food Infotech Magazine — July 2023
Read Article →Revolutionizing Food Safety: How BlockChain and Lighting Network are changing the game
Food Infotech Magazine — September 2023
Read Article →Revolutionizing Brain Tumor Detection: The CNN-based Medical Imaging Breakthrough
Informs London Magazine — December 2023
Read Article →Authored Books
Fundamentals of Convolutional Neural Networks with TensorFlow
A deep dive into Computer Vision, Object Detection, AI, CNNs, and TensorFlow.
Available on: Amazon, Kindle, Flipkart
Get The Book
Trust in the Age of Agentic AI Economy
A C-suite guide to building trustworthy, reliable agentic AI systems at scale. Drawn from the infrastructure OpenRAG Innovations has built in production — covering hallucination mitigation, retrieval reliability, multi-agent coordination, and the emerging trust frameworks enterprises need as AI systems become autonomous decision-makers.
Co-authored under OpenRAG Innovations Pvt. Ltd.
