Description:

The advent of ChatGPT has led to a huge boom in building GenAI & LLM powered chat interfaces across numerous sectors and domains. But building performant chatbots, especially in the customer facing segments come with a lot of challenges. In this talk, I’d like to share some of the challenges and common pitfalls we encounter while building customer facing chatbots. We hope that our experience would help members who are currently building or thinking of building chat assistants, in making better decisions and choices.

Brief outline of the talk:

• Current overview and common approaches to build chatbots (6 min)
• Basic architecture & advantages of Retrieval Augmentation & Generation (RAG) (3 min)
• Key challenges in building customer facing chatbots & ways to overcome them (15 min)
• Final architecture of RAG with best practices & conclusion (1 min)
• Q&A (5 min)

This talk mainly operates on high-level decisions and architecture for building chatbots and doesn’t go into implementation details. It’ll be a fun but informative session simplifying the key components in chatbot architectures with examples & scenarios. It is mainly suited for beginner and intermediate audience, and best suited for members who are currently building or thinking of building LLM powered chat assistants. Experience in software development is required to gain the most out of the session. Experience in machine learning is good to have but not mandatory.

Takeaways from the talk: At the end of the talk, takeaways for the audience would be,

• Understand various approaches for building chatbots and choose the ones best suited for their needs based on requirements & budget
• In-depth understanding of one such approach - Retrieval Augmentation & Generation (RAG) with architecture & key components
• Challenges and limitations with RAG based chatbots and ways to mitigate them
• Best practices to follow while designing customer centric chatbots with respect to security, performance and accuracy

Prerequisites:

• Experience in software development is required to gain the most out of the session
• Experience in ML & GenAI is good to have but not mandatory

Speaker Info:

Chaitanya is a technologist, engineer, entrepreneur and an angel investor. He is an alumnus of IIT Madras and currently the CTO of YNOS. He is also one of the creators of StartupGPT and an active open-source contributor, with his python libraries receiving over 120k+ downloads in total. Chaitanya has been working with Python for the past 7 years brings forth a unique perspective of design, product and tech in conjunction, enabling him to build creative and world-class products. He has a meticulous eye for detail and perceives that great things are a beautiful combination of small things built in a great way. He is minimalist, an amateur gamer and coffee lover. He is always up for discussing ideas, philosophy and pretty much anything about technology!

Speaker Links:

• LinkedIn - https://www.linkedin.com/in/chinnichaitanya/
• Github - https://github.com/chinnichaitanya
• Python Libraries - https://pypi.org/user/chinnichaitanya/
• StackOverflow - https://stackoverflow.com/users/3476748/chinni