AI Research & Publications
Advancing empirical AI research across media forensics, clinical NLP, and computer vision—bridging academic research across IITs with production safety.
IEEE Publication (2024)
Efficient Deepfake Detection using AI
Param Pandya, Collaborating Researchers — IEEE Advanced Engineering Systems and Practices Conference (AESPC)
This research addresses the critical challenges of deepfake detection within the field of media forensics, focusing specifically on improving model generalization across diverse datasets. To counter the domain shift caused by varying video compression formats, the study investigates reinforcement learning algorithms—specifically Deep Q-Networks (DQN) and Proximal Policy Optimization (PPO)—to dynamically optimize data augmentation strategies during model training. Utilizing robust deep learning architectures based on XceptionNet and InceptionResNetV2 backbones, the proposed framework adaptively learns generalized feature representations of visual manipulations. The integration of reinforcement learning helps the pipeline identify compression-resistant artifacts, mitigating performance drops on previously unseen datasets. This approach provides a trustworthy and robust methodology for visual media verification, contributing practical solutions for media forensics.
Research Interests
AI Systems & Large Language Models
Designing production-ready LLM systems with a focus on retrieval-augmented generation (RAG), autonomous multi-agent workflows, long-context reasoning, and building reliable architectures for real-world applications.
Generative AI for Healthcare
Developing safe and grounded clinical AI systems utilizing specialized biomedical language models, structured medical terminology integration, and constrained generation policies to improve clinical workflows.
Computer Vision & Media Forensics
Applying deep learning methodologies to visual understanding, medical diagnostics, deepfake media detection, model generalization, and explainable visual algorithms.
AI Systems Engineering
Engineering robust pipelines, highly scalable inference systems, vector databases, MLOps strategies, APIs, and containerized deployments for production AI environments.
Research Experience
IEEE Publication
Published research paper on Efficient Deepfake Detection using AI at IEEE AESPC 2024. Focus on reinforcement learning enhanced deepfake detection, cross-dataset evaluation, and media forensics.
Research Internship
Worked on deep learning experimentation, computer vision workflows, benchmarking, and AI model evaluation.
Research Internship
Conducted foundational AI research involving machine learning experimentation, data preprocessing, exploratory analysis, and academic research workflows.
Future Research Areas
Reliable Large Language Models & RAG
Researching systems to mitigate hallucinations in LLM workflows, optimizing multi-hop retrieval-augmented generation (RAG), and improving long-context reasoning in real-world environments.
AI Agents & Multi-Agent Systems
Exploring autonomous task-planning agents, tool-use integration, Model Context Protocol (MCP) standards, and cooperative multi-agent orchestration for complex workflows.
Healthcare AI & Clinical Decision Support
Adapting foundation clinical language models with ontology grounding (SNOMED CT, UMLS) to build transparent decision tools for diagnosis and prescription drafting.
Computer Vision & Medical Imaging
Developing trustworthy vision models for media forensics, deepfake detection, and medical imaging diagnostics utilizing self-supervised learning and explainable architectures.
AI Safety & Explainable AI (XAI)
Investigating model explainability (Grad-CAM, feature attribution), verifiable decision boundaries, and robust safety guardrails for deployment in critical systems.
Efficient Inference & Edge AI
Benchmarking and optimizing deep learning model deployment, quantization (INT8/FP4), pruning, and edge inference pipelines for resource-constrained client systems.
Curated Reading & Preprints
This repository features ongoing academic preprints, technical reports, and notes on foundational machine learning models.
Documents, technical slides, and reading logs are being indexed. This section will publish: