The Neuroblyx ecosystem brings together our models, hardware accelerators, and Cloud APIs.

Neuroblyx provides an ultra-low latency analog computing API for executing Vector-Matrix Multiplications and in-situ neural network training directly in Google Colab, Python, and ML production pipelines.

Generate Free API Key Google Colab Integration

One API Key Per Verified Email

Generate Your Neuroblyx API Key

Sign up with your email to receive a 6-digit verification code via Resend. Once verified, your unique private API key is generated immediately.

Integration Guide

How to Use Neuroblyx in Google Colab

Connect to the physical analog accelerator from Google Colab, Jupyter Notebooks, or any Python environment with 3 lines of code.

Fetch the Client SDK

Download ncc_client.py directly into your Google Colab instance.

Initialize with API Key

Pass your generated nbx_live_... API key and endpoint https://api.neuroblyx.in.

Dispatch Training / VMM

Call ncc.train(X, y) or ncc.vmm(vector, matrix) to execute analog computation.

Google_Colab_NCC_Quickstart.py

# ==========================================
# 🧠 RUN NEUROBLYX NCC IN GOOGLE COLAB
# ==========================================
# 1. Download official client SDK
!wget -q https://raw.githubusercontent.com/Rahulchaube1/ncc-product/main/ncc_client.py

# 2. Import & Initialize with your API Key
from ncc_client import NCC
import numpy as np

# Connect to Neuroblyx Analog Accelerator Cloud
ncc = NCC(
    api_key="YOUR_API_KEY_HERE",
    endpoint="https://api.neuroblyx.in"
)

# 3. Check Hardware Accelerator Status
print("🟢 Status:", ncc.status()["status"])

# 4. Train Any ML Model on Analog Crossbars
X = np.random.randn(200, 64)
y = np.random.randint(0, 10, size=200)

result = ncc.train(X_train=X, y_train=y, epochs=25, lr=0.02)
print("🎯 Final Accuracy:", result["final_train_accuracy"] * 100, "%")
print("⚡ Latency:", result["hardware_telemetry"]["training_duration_ms"], "ms")

Why Compute on NCC?

Non-Von Neumann Analog In-Memory Computing eliminates memory bus bottlenecks, allowing deep neural network layers to execute at physical Ohm's law relaxation time.

Non-Von Neumann Fabric

Analog Crossbar Architecture

Synaptic weights stored as physical conductances (G_ij in Siemens), eliminating high-bandwidth memory shuttling.

1,700x Energy Advantage

Operates at sub-milliwatt power dissipation per crossbar tile compared to 700W TDP of conventional digital GPUs.

0.004 ms VMM Latency

Continuous physical current accumulation across column bitlines in constant $O(1)$ time.

REST Specifications

Official API Endpoints

Endpoint	Description	Latency	Auth
POST /api/v1/vmm	Analog Vector-Matrix Multiplication (I = V · G)	0.004 ms	Bearer Key
POST /api/v1/train	In-Situ Model Training on Analog Crossbars	In-Situ	Bearer Key
GET /api/v1/status	Accelerator Health, Mesh Cores & Thermal Drift	1.2 ms	Public
POST /api/v1/auth/signup	User Signup with Resend OTP Verification	Email	Public