I build the infrastructure layer between intelligence and execution — the systems that let AI agents operate in production without becoming a liability.
My work sits at the intersection of AI security, runtime governance, and operational memory: making autonomous systems that are safe to deploy, auditable by design, and provably compliant.
|
AI Agent Firewall & Governance Runtime security layer for autonomous AI agents. Every tool call is intercepted, scanned through a 7-layer firewall, evaluated against policy rules, and either allowed, blocked, or escalated to a human — before it executes.
|
Operational Memory for LLMs Persistent, structured memory engine that makes AI agents learn from experience. Captures events, extracts lessons via LLM reflection, and retrieves the most relevant context before each new task — reducing from-scratch reasoning and cutting inference costs by ~43%.
|
Verified AI Connectivity Secure identity and communication layer for AI-to-AI and human-to-AI interactions. Built on the principle that trust in agentic systems must be earned, cryptographically verifiable, and auditable at every hop in the chain.
|
Languages Python · TypeScript · SQL
AI / ML LangGraph · CrewAI · PydanticAI · OpenAI · Anthropic Claude
Databases MongoDB · PostgreSQL + pgvector · Redis
Backend FastAPI · Beanie ODM · ARQ workers · asyncio
Frontend React · Vite · Tailwind CSS · Cloudflare Workers
Infra Railway · Vercel · Docker · GitHub Actions
Security AES-256 · SHA-256 hash chains · bcrypt · JWT · OWASP
Compliance SOC 2 · GDPR · HIPAA · EU AI Act · OWASP Agentic Top 10
Protocols MCP · REST · WebSocket · pub/sub (Redis + in-process)
AI Agent Security Prompt Injection Defense LLM Governance
Operational Memory Multi-Agent Systems Zero Trust AI
Compliance Automation Human-in-the-Loop Design Agentic AI Safety
Runtime Policy Engines Audit-by-Default Architectures LLMOps
| 📧 | vanawalanishant@gmail.com |
| 🌐 | polaxis.io |
| 📚 | docs.polaxis.io |
| 💼 | Enterprise: sales@polaxis.io |
"The question isn't whether AI agents will be powerful enough to matter. It's whether they'll be safe enough to trust."
