Microservices, Amazon Titan and More.

Welcome to the Inaugural edition of “Tech??vara”.

This is a fortnightly update on the latest?technology trends and topics from the Office of the CTO of Ganit. The word?“?vara”?in Indian Classical music refers to the seven building blocks of music - Sa, Re, Ga, Ma, Pa,?Dha?and Ni – from which Ragas are composed – and this newsletter aims to breakdown a larger trends or insights in the?tech?industry into smaller blocks that are easier to understand.?

Over the years several key?technology trends have stood the test of time and have gradually become the de-facto standard in their field. This week you can expect

• The Microservices architecture series
• Security 101
• Amazon’s official entry in the Generative AI race
• the “power of showing up” - an important lesson in customer success

So, let's dive in!

The Basics: The advent of the Internet and the exponential scale up of volume and complexity of Internet scale applications rendered several traditional approaches in?technology obsolete. The monolith architecture was one such artifact that gave way to a decomposed architecture based on Microservices. The fundamental building block of this approach is a Microservice which are:

• Loosely coupled
• Organized around a business domain or capabilities
• Independently deployable
• Owned by a small team (also referred to as the “Two-Pizza Rule” a term popularized by Amazon)

While there is no standard that covers Microservices architecture, The article by Martin Fowler listed above and the book - Microservices Design Pattern (along with its author site) are the most definite sources on this topic. We will cover the key design patterns for Microservices architecture in the upcoming parts of this series.

The Basics: The Software Supply chain attack has fast emerged as a key vector for security breach in recent times. This attack is executed by compromising a common library or system that is indirectly used by organizations.

A widely reported example of this is the SolarWinds breach?that impacted reputed organizations like Microsoft, FireEye, the Pentagon, Homeland Security, and went undetected for months. Other attacks include the Log4J compromise. SDKs/Libraries are especially vulnerable to such attacks due to their large install base and dependence on third party libraries that may not always be detected.

We recommend using tools such as Dependabot?and Bandit?to scan your code for security vulnerabilities including those found in out dated libraries.

With the Generative AI space heating up it was only a matter of time before AWS jumps in with its own foundational model. That’s exactly what they did with their announcement of Titan.

Titan Text focused on summarization and text generation and Titan Embeddings designed to apply LLMs to personalization and search domains have been trained on the vast dataset generated on Amazon.com. Another important feature that didn’t get as much coverage is the Data escrow model that AWS has adopted for enabling custom training of LLMs. This ensures that none of the customer data leaves their VPC and its not used as a training set for the public LLM itself.