Rechat AI Innovators: Engineer Adam Schwartz
This week we have the pleasure of connecting with Adam Schwartz, a Rechat senior software engineer based in Budapest with over 12 years of experience. He holds a master’s in computer science, and his fascination with machine learning was sparked while pursuing his academic studies. The rise of products like ChatGPT has accelerated his work, and he said he’s excited to contribute to projects like Lucy at Rechat, using AI to empower agents and brokers.
What brought you to Rechat and what intrigued you about your role/work here?
My journey to Rechat unfolded in a rather serendipitous way. I was leading the tech team at a real estate agency in New York, where we developed an in-house product for agents. However, due to circumstances, the tech team had to be laid off. It was during this time that the CEO of my former company, who had a connection with Shayan (Rechat CEO), reached out to explore opportunities. After a series of conversations with Abbas, the leader of the backend team at Rechat, where we discussed our experiences and goals, I saw a great alignment. The opportunity to contribute to an innovative company like Rechat, combined with the chance to work on meaningful projects, influenced my choice to become part of the Rechat team.
What is your title and what does your job look like on a day to day basis?
As a senior software engineer on the backend team, my day-to-day revolves around bringing product ideas to life by diving into the code. I spend my time exploring the existing codebase, identifying areas for improvement, and collaborating with team members. Whether I’m working on tasks assigned by Emil (Rechat CTO) or addressing requests from the product team, my role is all about transforming concepts into tangible code and contributing to the continuous enhancement of our projects.
What has been your role as it relates to the development of Lucy?
In the initial stages, I actively participated in the planning phase, exploring tools and ultimately selecting LangChain to work with. I took charge of implementing the foundational features of Lucy, focusing on building a seamless chat experience complete with history.
As the development progressed, I delved into the architecture of tools, enabling Lucy to read and manipulate data within users’ Rechat accounts. This involved empowering Lucy to perform tasks such as creating new contacts in their CRM or conducting searches. It’s been a dynamic journey, from laying the groundwork for Lucy’s functionality to ensuring users can leverage its capabilities effectively within their Rechat ecosystem.
Would you be able to explain a bit about the process of training and fine tuning as it relates to Lucy, in simple terms?
Fine-tuning Lucy is like coaching a friend who’s learning to respond in different situations. In this process, we generate test conversations, purposely including scenarios where Lucy might make errors. We collect and tag these errors, then manually adjust and refine Lucy’s responses based on these corrections, ensuring that it learns from its mistakes. This ongoing fine-tuning process enables Lucy to continually improve, delivering responses that are increasingly accurate and contextually relevant as it learns from its experiences.
Am I right in thinking that this work is what will ultimately make Lucy a more exceptional real estate AI tool than most?
Absolutely! The dedicated effort we put into fine-tuning Lucy is what sets it apart. By consistently refining its understanding, addressing errors, and adapting to various scenarios, we aim to ensure that Lucy not only meets but exceeds expectations. This commitment to improvement is what distinguishes Lucy, making it a more sophisticated and effective tool over time.
What is the most exciting part of working in AI?
The most exciting part of being in the AI realm is watching its transformative impact on tech and our daily lives. It’s like navigating uncharted territories, working on complex problems, and contributing to innovations that push boundaries. AI allows us to craft new solutions that evolve, offering endless opportunities for creativity and making a real difference in how we interact with and benefit from technology.
What is the most challenging part of working in AI?
The most challenging aspects of working in AI, for me, revolve around the dynamic nature of the field. Keeping up with the latest trends and adapting to ever-changing tech can be quite a challenge. Dealing with the unpredictable nature of AI can be a bit of a challenge, too, as occasionally the outcomes are unpredictable, and we have limited control over certain aspects.
How does Rechat keep agents’ data safe?
Lucy, powered by OpenAI’s ChatGPT, relies on OpenAI’s robust data safety measures in the background. Additionally, the tools Lucy utilizes to retrieve and modify data from Rechat’s database are designed with user privacy in mind—ensuring that only the user’s data, which is interacting with Lucy, is accessible to them. It’s worth noting that, for the fine-tuning process, we exclusively use conversations generated with Lucy, ensuring the confidentiality of agents’ data throughout this refinement process.
