‍

At WITH we invests over $1.2M in building ChatFAQ, the new generation of FAQ chatbots. The role of the NLP engineer (Natural Language Processing) becomes critical to the project's success. The NLP engineer is responsible for designing and implementing the NLP engine, the cornerstone capability to understand and interpret users input in the chat experience.

NLP & ML Challenges in Chatbot Development

One of the key challenges faced by the NLP engineer is the fact that users often make grammatical or orthographic mistakes, use abbreviations, or employ domain-specific vocabulary that the chatbot struggles to recognize. To address these challenges, the NLP engineer must develop, adapt and train the NLP engine, using robust datasets that take into account various user inputs, business scenarios and multilingual specificities. This is an iterative process done in close collaboration with the rest of the development team.

Another challenge is the need for the chatbot to adapt to users' needs and provide relevant responses to their questions as the conversation unfolds. The NLP engine has to analyze user intents and extract accurate and appropriate responses based on the context of the conversation. This requires a deep understanding of the business domain and the ability to develop and maintain a comprehensive knowledge database.

Open-Source Solution for Chatbot Maintainability and Adaptability

To address these challenges, WITH is building an open-source model that will give developers control over the maintainability and adaptability of the chatbot under live conditions. The NLP engineer ensures that the NLP engine is integrated seamlessly into the chatbot infrastructure with proper operational tools in place to understand, measure and tweak its behaviour in order to achieve the expected accuracy and performance KPIs.

The success of ChatFAQ, our new generation of FAQ chatbots project, depends on the NLP engineer's ability to develop and implement this powerful and maintainable NLP engine that goes beyond simply understanding and interpreting natural language. WITH's investment in this project is a testament to the importance of making NLP a maintainable and adaptable component of the NLP technology stack to provide highly accurate service to end-users.

A Day in the Life of an NLP Engineer

A normal day for an NLP engineer is spent exploring the latest developments in generative and non-generative language models, in the area of quality and optimization, to make these large models viable. They work with robust datasets to train the NLP engine and measure their effectiveness against grammatical or orthographic mistakes, abbreviations, and domain-specific vocabulary.

The NLP engineer drives and assists the Machine Learning Operations (MLOps) architecture requirements to ensure that future customers have the best operational quality of their chatbots, including optimizing all essential components of #ChatFAQ and even participating in the solution of certain development and design issues as well.

It is extremely exciting to see how the most modern language models can be put to real productive use and not remain mere research, and how they can contribute to the success of projects like ChatFAQ. This is a unique opportunity to influence the Conversational AI roadmap while still working for an AI open source project and a company like WITH, a great example of a disruptive company that is always committed to the latest technological advances.

Working as an NLP engineer on ChatFAQ represents a unique opportunity to work on a large and complicated project with all the challenges that come with it, including engineering, optimization, and necessary infrastructure. It also allows to stay up to date with the latest developments in research, as few jobs are like this. Additionally, working on a product for all kinds of companies requires taking into account many perspectives in which the product will be used.

Apart from the technical aspects, it is also necessary to communicate with customers and other sectors to align all parties involved and create a successful product. Overall, it is a challenging but rewarding experience that provides a chance to work with cutting-edge technology and contribute to the development of conversational AI in a meaningful way.

Large Language Models and controllable outcomes are exciting prospects in the field of conversational interfaces. There is a big future for generative AI and its impact on the internet. Although we are still in the early stages of these models, we are excited about how With is positioned to leverage the best of what is yet to come in this space. The most engineering part of these models is to make them as efficient and maintainable as possible, using techniques such as hardware optimization, user-friendly MLOps pipelines and other tricks, wink wink.

Stay tuned!

‍

‍