ErenLabsErenLabs
Research
Active research streams at ErenLabs cover continuous learning systems, the IGRIS architecture, embedded telemetry, and intelligence engineering. This page tracks both long-form research articles and individual log entries.
ErenLabs operates as an independent research lab focused on building AI systems that keep learning after they are deployed. Most of the work sits at the intersection of continuous learning, embedded engineering, and telemetry infrastructure — the unglamorous plumbing that has to exist before a learning agent can run in the real world.
Two threads are documented in long form below. Shorter weekly logs follow underneath. The accompanying video series is published on the IGRIS YouTube channel.
Research Articles
IGRIS — Continuous Learning Architecture for Embodied AI
A technical walkthrough of the IGRIS runtime: how the learning core, telemetry pipeline, and cognitive execution bridge fit together so an embodied agent can learn from real-world signals continuously, not just at training time.
Read article → ConceptWhat is Continuous Learning?
A plain-English explainer covering continuous learning, online learning, and lifelong learning — how they differ from offline training, the hard problems (catastrophic forgetting, drift), and where the idea actually matters.
Read article →Active Research Logs
IGRIS — Continuous Learning Architecture v1 Design
First-revision design for the learning runtime that backs the IGRIS agent. Covers selective updates, bounded drift, and the sliding-window memory structure that lets an embodied agent learn without catastrophic forgetting. Full writeup in the IGRIS article.
Multi-Stage Telemetry Pipeline — CubeSat Research
Multi-rate, lossy-transport-tolerant telemetry pipeline designed around CubeSat-class radio links. Handles multi-sensor timestamp alignment, sensor health channels, and graceful degradation when packets are lost.
Real-Time Control System — DAQ Unit Firmware
Microcontroller firmware for the DAQ (Data Acquisition) unit that sits closest to the sensors. Provides hardware-timestamped, confidence-tagged readings to the higher-level telemetry pipeline at 1 kHz IMU rates.
Cognitive Execution Bridge — Agent-to-Device Interface
The translation layer between learning-runtime decisions and physical device commands. Enforces per-device rate limits, confidence thresholds for human-in-the-loop deferral, and full action auditing.
ErenLabs API Infrastructure — Cloudflare Tunnel Setup
Public API endpoint at api.erenlabs.com backed by a Cloudflare Tunnel so the lab can expose telemetry and health endpoints without opening inbound ports on a home network.
Research in Public
Follow along with active research on the IGRIS YouTube channel.