Hi, I'm Siddhant Khare 👋

📍 India | 🛠️ Engineer @ Ona (formerly Gitpod) | 🔐 OpenFGA Core Maintainer | 🎤 International Speaker | siddhantkhare.com

Building infrastructure for AI agents - context efficiency, least-privilege security, production-grade tooling.

Current Projects

Agent Infrastructure & Context Engineering

• 🧬 Distill - Deterministic context deduplication for LLMs. Clean context in ~12ms, zero LLM calls.
• 🧪 ContextLab - Open-source LLM context engineering toolkit: analyze, compress, visualize.
• 💾 TokenVM - High-performance runtime treating LLM KV cache as virtual memory with page-based eviction.
• 📊 KV-Cache Profiler - Profile LLM GPU memory needs before deployment, not after.
• 🔬 LLMTraceFX - GPU-level LLM inference profiler with kernel timing and AI-powered bottleneck detection.
• 🤖 Agentflow - Kubernetes for AI agents - orchestration runtime, prompt ops, security layer, observability, and cost-aware scheduling.
• 🌐 AI Agent Orchestrator - Multi-agent AI system on Cloudflare Workers + Containers.
• 🔢 LLM Parallelism Explorer - Research tool for optimizing parallelism strategies in LLMs (MoE focus).
• 💰 CloudArb - GPU arbitrage platform for AI compute optimization.

Agent Security & Authorization

• 🔐 agentic-authz - OpenFGA + MCP authorization gateway for AI agents. Fine-grained access control at team, project, and tool levels.

• 🔏 Agentic Authorization - Authorization patterns for autonomous AI agent systems using ReBAC with OpenFGA.

• 🔗 A2AS Implementation - Proof-of-concept of Agent-to-Agent Security framework.

• 🛡️ Sentinel AI - Hermetic CLI for security scanning and dead-code detection with LLM-powered triage.

• 🔒 actionsec - Fast, local-first CLI for GitHub Actions security analysis.

• Developer Tools & MCP Servers

• Applications

• Legacy / Earlier Work

• What I'm Doing

• Recognition

• Research Focus