Master Designing, Integrating & Deploying Enterprise AI Apps

Explore the difference between io bound and cpu bound tasks and distinguish concurrency from parallelism, illustrating how cpus, cores, and scheduling shape asynchronous io in enterprise ai apps.

Explore building concurrent applications with threads, using event loops, queues, and call stacks. Compare sequential and thread-based execution, and note overhead and the shift to event-driven approaches.

Master concurrency using event loops and async programming in Python, scheduling tasks on the call stack while other tasks wait for IO, using create_task to manage concurrent execution.

Explore NATS, a high-performance cloud-native open-source message bus that enables publish-subscribe with subject-based routing, distributed clustering, and authentication and authorization for secure, language-agnostic apps.

Learn how to download and run the nats-server from the open source project, using prebuilt binaries across Linux, Mac, and Windows, with optional port changes, tracing, authentication, TLS, and clustering.

Develop two simple apps that communicate via NATS, with a publisher and a subscriber, evolving through three versions that cover APIs, object-oriented programming, and state management.

Publishers and subscribers in version one use an async io loop and the next client library to connect to a net server, subscribe to topic one, and handle messages.

Explore version two of the publisher and subscriber using NATS and object-oriented patterns, introducing a dedicated publisher class with connect and start methods, and subscriber callbacks as class methods.

Explains version 3 of the publisher using NATS and threads to publish messages every second, with a data receiver and a publisher object, guided by the single responsibility principle.

Cluster two net servers to connect publisher and subscriber across machines, validate with ping pong, and route messages only to active subscribers.

Discover FlatBuffers, a Google binary data format for high-performance apps, offering compact, memory-efficient serialization with in-place deserialization and optional field support across languages.

Install flag before compiler and libraries for Windows, macOS, and Linux. Generate language stubs from a schema to enable Python, C++, C#, and JavaScript interoperability.

Explore the essentials of schema syntax and usage in Python, with a multi-part, step-by-step approach that updates your code as features are added or removed.

Explore primitives, strings, and tables in the flag before schema language, and learn Python serialization of robot objects using a builder and offsets.

Model robot power levels with enums in a byte-backed primitive type, defining three states: starter, decent, and top notch. Add the power level property and serialize strongly typed data.

Embed a custom object in a table by introducing a manufacturer table, importing its type, and building robot objects via a builder with manufacturer data and location for serialization.

Learn to design multi-file schemas by splitting robot and person tables into separate files, using include to resolve cross-file references, and updating compiler flags.

Explore serializing a vector of custom objects by embedding a robots vector inside a person object, including markers, offsets, and builder patterns with a Python sample walkthrough.

Build and deploy an enterprise ai app pipeline that classifies human activities from sensor data using three apps (stream, classifier, decided), a test data file, and a shared flat protocol.

Design and serialize flatbuffer messages for enterprise ai apps in python, including a classification request with a unique id and base64 sensor data, flattening 2d data for transmission.

Explore the streamer app's publisher, sensor data handling, and endless loop that shuffles test data and sends serialized examples through an event loop.

Learn to build a classifier app that uses the Janus thread-safe async queue to pass classification requests between main and async threads, with NATS messaging and top predictions.

Subscribe to the decision maker and evaluate classifier predictions against a 0.9 threshold to trigger an action when the highest prediction meets the threshold; otherwise, report inability to decide.

Run the full four-part AI pipeline locally and on an edge device by launching the streamer, classifier, and decider in a clustered mode with a virtual environment.

Docker uses Linux namespaces and cgroups to isolate apps in containers, built from images and managed by a lightweight container engine with registry support.

Pull and run simple docker images from the public registry, explore ubuntu and python images, run containers interactively, and manage images and containers with list and remove commands.

Create a custom Docker image by writing a Dockerfile with a base image and tags, set environment variables, define a working directory, copy requirements, and expose ports. Build and run the image with Docker build and Docker run, using caching and tagging to manage versions.

Write dockerfiles for the streamer, classifier, and decider, using python 3.8 buster as the base, install requirements, set environment variables, and push images to a Docker registry.

Run and connect docker containers for NATS, streamer, classifier, and decider by mapping host ports to container ports, enabling inter-container communication and exploring dockerized workflows.

Learn how Docker Compose orchestrates multi-container apps with a YAML file that defines services, images, ports, and dependencies, then use up and down to run and clean up.