++++

Atharva
Mendhulkar.

Porygon

Systems Engineer & AI Researcher. Exploring the intersection of physics-informed machine learning and kernel-level infrastructure.

AI Systems

CAPS

github

A deterministic payment kernel designed to create a trust boundary between AI reasoning and payment execution.

4 layers
Security gates
Sub-50ms
Latency
Full audit trail
Explainability
Deterministic
Decision Logic
Agentic FinanceSecurityAI GovernanceDeterministic Systems
Open Source · Research Prototype
AI Systems

AVARA

github

Runtime security and governance layer for autonomous AI agents — enforces intent, permissions, provenance, and safety at execution time, not training time.

8 guards
Security layers
< 5ms
Runtime overhead
100%
Audit replayable
3 lines
LangChain Integration
AI SecurityAgent GovernanceRuntime SafetyLLM
Open Source · Production-Ready
Systems

K-PHD

github

Embeds into the Linux scheduler hot-path to measure nanosecond-level process wait times and predict CPU starvation before applications hang.

nanosecond
Granularity
1,024 per-PID
Hash buckets
2ms / 5ms
Warn/Danger
GPL v2
License
Kernel ModuleTracepointsSystemsPredictive
Linux Kernel · GPL v2
SaaS

Floework

github

Replaces time-tracking with behavioral signal analysis — models the causal chain Focus → Effort → Task Progress → Team Outcome.

7 signals
Execution signals
WebSocket
Real-time
Multi-tenant
Architecture
Cloud-native
Deployment
SaaSProductivityReal-TimeAnalytics
Live · floework.vercel.app
++++
The Knowledge Garden
An interactive knowledge graph built from research notes. Synthesizing 20+ nodes across fluid dynamics, kernel architecture, and agentic security.