Peer-Reviewed Research
Research
7 peer-reviewed publications across AI agent reliability. Mathematical foundations for reliable AI agents.
Publications
Formal methods and mathematical guarantees for every layer of the AI agent stack.
Agent Behavioral Contracts: Formal Specification and Runtime Enforcement for Reliable Autonomous AI Agents
agentAssert · Research Project
We present Agent Behavioral Contracts (ABC), a formal framework that combines design-by-contract principles with stochastic process theory to provide runtime behavioral guarantees for autonomous AI agents. ABC introduces a four-component contract structure {P, I, G, R} with mathematical drift bounds via Lyapunov stability analysis, achieving Θ=0.9541 aggregate compliance across 200 benchmark scenarios.
SkillFortify: Formal Verification Framework for Agent Skill Supply Chain Security
SkillFortify · Open Source
A formal verification framework that provides mathematical guarantees for agent skill security across 22 frameworks. Achieves F1=96.95% detection accuracy with zero false positives. SkillFortify addresses the growing supply chain risk in AI agent ecosystems by verifying skill integrity before execution, preventing prompt injection, data exfiltration, and privilege escalation.
AgentAssay: Token-Efficient Regression Testing for Non-Deterministic AI Agent Workflows
AgentAssay · Open Source
A token-efficient stochastic testing framework for non-deterministic AI agent workflows. AgentAssay introduces regression testing methodologies that account for the inherent non-determinism in LLM-powered agents, enabling reliable CI/CD pipelines for agent systems with minimal token consumption.
SuperLocalMemory: Privacy-Preserving Multi-Agent Memory Architecture
SuperLocalMemory · Open Source
A local-first memory architecture for AI agents that preserves privacy while enabling multi-agent memory sharing. Zero cloud dependency, works with Claude Code, Cursor, ChatGPT, and Perplexity. SuperLocalMemory provides persistent context across sessions with full data sovereignty, ensuring no sensitive information leaves the developer's machine.
SuperLocalMemory V3.3: Cognitive Memory Architecture with Adaptive Retrieval and Neural Consolidation
SuperLocalMemory V3.3 · Open Source
Introduces cognitive memory architecture with adaptive retrieval channels and neural consolidation for AI agent memory systems. Extends SuperLocalMemory with multi-channel retrieval, Fisher-Rao importance scoring, and autonomous memory lifecycle management for persistent, privacy-preserving agent context across sessions.
SuperLocalMemory V3: Scalable Privacy-Preserving Memory for Multi-Agent AI Systems
SuperLocalMemory V3 · Open Source
Presents the V3 architecture of SuperLocalMemory with scalable, privacy-preserving memory for multi-agent AI systems. Introduces semantic search, session-aware retrieval, and zero-cloud dependency memory management enabling full data sovereignty for AI agent workflows.
Qualixar OS: A Universal Agent Operating System with Type-C Command Protocol
Qualixar OS · Open Source
Introduces Qualixar OS, a universal agent operating system with the Type-C command protocol for AI agent orchestration. Provides a unified interface for agent lifecycle management, multi-topology execution, and cross-transport communication enabling any AI agent to run on any platform through a single standardized port.
Key Results
Rigorously evaluated across diverse enterprise domains.
Domain Performance
Benchmark results across 7 real-world enterprise domains.
Academic Profiles
Follow ongoing research and publications.
Research Interests
Building the mathematical and systems foundations for trustworthy autonomous AI.
Cite This Work
BibTeX entries for these research publications.
@article{bhardwaj2026abc,
title={Agent Behavioral Contracts: Formal Specification and
Runtime Enforcement for Reliable Autonomous AI Agents},
author={Bhardwaj, Varun Pratap},
journal={arXiv preprint arXiv:2602.22302},
year={2026},
url={https://arxiv.org/abs/2602.22302},
doi={10.5281/zenodo.18775393}
}@article{bhardwaj2026skillfortify,
title={SkillFortify: Formal Verification Framework for
Agent Skill Supply Chain Security},
author={Bhardwaj, Varun Pratap},
journal={arXiv preprint arXiv:2603.00195},
year={2026},
url={https://arxiv.org/abs/2603.00195},
doi={10.5281/zenodo.18787663}
}@article{bhardwaj2026agentassay,
title={AgentAssay: Token-Efficient Regression Testing for
Non-Deterministic AI Agent Workflows},
author={Bhardwaj, Varun Pratap},
journal={arXiv preprint arXiv:2603.02601},
year={2026},
url={https://arxiv.org/abs/2603.02601},
doi={10.5281/zenodo.18842011}
}@article{bhardwaj2026slm,
title={SuperLocalMemory: Privacy-Preserving Multi-Agent
Memory Architecture},
author={Bhardwaj, Varun Pratap},
journal={arXiv preprint arXiv:2603.02240},
year={2026},
url={https://arxiv.org/abs/2603.02240},
doi={10.5281/zenodo.18709670}
}