I'm Vishal S. Pandey, an Applied AI Research Engineer with a dedicated focus on the evolving landscape of Artificial Intelligence. My professional trajectory is defined by a deep-seated commitment to translating cutting-edge AI research into practical, impactful systems. I am particularly invested in advancing large language models (LLMs), developing innovative optimization strategies, and constructing robust multi-agent architectures.
My research mission is centered on addressing critical challenges in AI, including the intricate aspects of AI safety and control. This involves exploring the mathematical foundations of AI, designing novel architectures, and rigorously evaluating their efficacy and ethical implications. I am driven by the pursuit of building autonomous systems that are not only highly performant but also inherently reliable and aligned with human values
My current work encompasses the independent development of research-grade ML systems and toolkits, often leading to open-source contributions that foster broader collaboration within the AI community. Notable projects include:
- AgenticCyberOps: A LangGraph + LangChain multi-agent framework engineered for autonomous cybersecurity operations, with a focus on AI control and information security of safety-critical AI systems.
- AgroSense: A Streamlit-deployed tool leveraging Vision Transformers and data pipelines for soil classification (94.3% accuracy) and crop recommendation (97-98% accuracy).
- Optimization for LLMs (work in progress): Ongoing research exploring optimizer hybrids for resource-efficient fine-tuning of transformers, achieving a 15% perplexity reduction on distilled GPT-2.
- Extended Decision Transformer (work in progress): Design of an energy-efficient AI system for long-horizon environments, demonstrating a 2x reduction in inference energy consumption with performance parity on Gym tasks.
- Diabetes Types Prediction: Development and evaluation of machine learning models for diabetes classification and subtype prediction on the Pima dataset.
Educational Repositories
I've developed comprehensive guides on Machine Learning and Deep Learning, available at
Note: As my latest projects are currently under wraps and undergoing active development, I will provide comprehensive details upon their official publication.
You can find my code and contributions on my GitHub and Kaggle profiles. For updates on my research, thoughts on AI, and general musings, feel free to follow me on X (formerly Twitter).