TensorFlow / Keras

What it is

Google's deep learning framework. Keras (the high-level API) is one of the friendliest ways to build models; TFLite is the king of mobile / edge deployment for Android.

How Vaaani uses it

• On-device inference for custom Android AI tools
• Production serving via TF-Serving with batched gRPC
• TensorFlow.js for in-browser AI features
• Legacy enterprise codebases that already run on TF

Why it makes the cut

When the AI worker has to ship inside a customer's Android app, TFLite is unmatched. PyTorch Mobile is catching up; TF still wins for low-power edge today.

Sample code

import tensorflow as tf

model = tf.keras.Sequential([
    tf.keras.layers.Dense(128, activation="relu"),
    tf.keras.layers.Dense(10, activation="softmax"),
])
model.compile(optimizer="adam", loss="categorical_crossentropy")