Product & company site: enotech.systems
Manage temporary power infrastructure for large-scale temporary operations.
Architected a 10,000-node relational database to track asset placement and resource consumption.
Reduced fuel consumption by 10% and optimized resource allocation through data-driven decision making.
Design sovereign edge compute for kinetic, off-grid environments: local persistence, store-and-forward telemetry, and minimal reliance on continuous cloud connectivity.
DDIL (Denied, Disrupted, Intermittent, Limited). High thermal variance. Variable power input (Solar/DC).
Architecture emphasizes hardware watchdog recovery, power-aware orchestration, and durable local logging: exercised in mobile test-rig and lab scenarios, not published production SLAs.
Deploying persistent AI agents and data services in "Denied, Disrupted, Intermittent, and Limited" (DDIL) environments presents two critical failures:
Goal: Build a "Zero-Dependency" node capable of running local LLMs and agent swarms with <100W power draw.
The VITO Node utilizes a hardware-first approach to resilience. Unlike cloud clusters, the physical layer is tightly coupled with the logic layer.
The VITO Node v4.0 is a self-contained infrastructure stack designed to replace fragile cloud dependencies with robust local sovereignty. It integrates a Cognitive Core (Ollama/CrewAI) for autonomous decision making and a Hardware Watchdog layer for physical resilience.
Power Layer: 2kW Solar Array feeding 5kWh LiFePO4 buffer. Direct DC-to-DC regulation avoids inverter efficiency loss.
Compute Kernel: Linux Docker Host with Docker Compose orchestration. Selected for deterministic restart policies and low overhead compared to Kubernetes (critical for solar budget). Custom hardware watchdog monitors voltage at the solar controller.
Cognitive Core: Ollama inference engine running local LLM models (granite4:3b, qwen2.5:3b) with CrewAI orchestrating autonomous agent workforce. Vector Memory (ChromaDB) enables persistent agent memory across restarts.
Watchdog Protocol: Voltage sensors monitor battery SoC. If battery drops <20%, the system gracefully terminates non-essential Docker containers before hard power loss. Hardware watchdog timer executes hard reset if software hangs.
Storage Layer: NVMe SSD with atomic write guarantees. Immutable logs and a store-and-forward buffer are designed to preserve telemetry across network outages. CrewAI run history stored locally for audit and recovery.
Network & Observability: Zero-trust mesh (Tailscale) for secure remote access. LoRa telemetry for low-power long-range communication. Datadog metrics and Sentry error tracking for operational visibility.
Power-aware orchestration script that manages container lifecycles based on telemetry:
def power_governor(telemetry):
"""
Adjusts compute load based on battery telemetry.
Prioritizes critical infra (databases) over
expansive logic (LLMs).
"""
battery_soc = telemetry.get('soc')
solar_input = telemetry.get('pv_watts')
if battery_soc < 20:
print("[CRITICAL] Power Critical. Shedding Load.")
docker_client.containers.get('llm_inference').stop()
docker_client.containers.get('agent_swarm').stop()
elif battery_soc > 80 and solar_input > 500:
print("[NOMINAL] Solar Surplus. Resuming Training Jobs.")
docker_client.containers.get('llm_train').start()Current product specs and procurement: enotech.systems/vito
Docker Orchestration vs. Kubernetes: Selected Docker Compose for power overhead. Kubernetes control plane consumes 20-30W minimum. On a solar budget, every watt matters. Docker Compose provides deterministic restart policies without the orchestration overhead.
Direct DC Power: Eliminated AC inverter layer to reduce conversion losses. Server PSU operates directly from LiFePO4 battery bank with buck converters for voltage regulation. Saves ~15% power compared to AC-inverter-AC path.
Hardware Watchdog: Software can fail silently during brownouts. Hardware watchdog timer monitors kernel heartbeat. On hang detection, executes hard power cycle autonomously. Enables "Zero-Touch" recovery without human intervention.
Rapid deployment of DC Fast Charging infrastructure.
Managed utility coordination, supply chain logistics, and contractor execution for high-voltage grid ties.
Successful deployment of critical charging nodes on the national network.