Today we’re catching up with Rechat’s AI Advisor, Hamel Husain. He’s been working on machine learning for 20+ years, and has implemented AI products in a wide variety of industries such as telecommunications, entertainment, restaurants, retail and technology.
He’s also spent a fair amount of time building open source tools for other data scientists and machine learning engineers, which you read about in his blog. He has also worked on language models for an extensive period of time, including work he led while at GitHub called CodeSearchNet. More recently, he’s started Parlance Labs, a consultancy that helps people with large language models, particularly around evaluation, fine-tuning and serving.
What brought you to Rechat and what intrigued you about your role and work here?
Hamel: Rechat represents a canonical example of the type of company LLMs are supposed to benefit the most: a scrappy company that doesn’t have an army of machine learning specialists. When I saw the Rechat product and the way the team was thinking about LLMs I knew they had the right environment for a huge success with AI: a product that already works well without AI, good engineering practices, and a growth mindset. I knew that Rechat’s success could serve as an example for the whole industry, so I was excited to be involved in driving Rechat towards effective AI.
What has been your role as it relates to the development of Lucy?
Hamel: I’m advising on all aspects of Rechat’s efforts to create the industry’s first real estate agent co-pilot, Lucy. At this stage in Lucy’s development, I’ve been focusing most on evaluation, fine-tuning and DevOps. However, as development on Lucy progresses, my focus may shift to other aspects.
By investing in an evaluation and data curation framework, we drastically increased our development velocity on AI. These tools allow us to get immediate feedback when we make changes or improvements to our AI system. This has resulted in dramatic improvements to our AI system over a short period of time. These foundations allow us to feel confident about shipping AI that is delightful to customers in the near future.
What does it mean to be an “AI Advisor” in the most basic, practical sense?
Hamel: My job is to help people be more successful with AI. My attitude is to do whatever is necessary to accelerate my clients, which requires thinking about DevOps, software engineering, product, UX, and more traditional machine learning items like data engineering, evaluation, metrics and fine-tuning/training.
A good advisor listens to what their customer is trying to accomplish and thinks holistically about what is needed to make them successful, even if that diverges from what they think they want.
You’ve only recently started working with Rechat. What has been your experience so far?
Hamel: Rechat is the blueprint of a company I would love to start one day:
- Remote first, global workforce
- Run very efficiently both financially and from a people management perspective
- Excellent company culture
- Focus on simplicity while still maintaining engineering excellence. Homegrown solutions are not built unless it is really necessary.
This focus on minimalism and simplicity is a breath of fresh air relative to your typical VC-funded silicon valley company. This is because a lot more gets done with fewer resources. Seeing companies like Rechat become successful with AI finally gives credence to the term “democratizing AI.”
Do you think real estate agents will be excited to adopt these new tools as they become available or do you think agent skepticism toward AI may hinder adoption?
Hamel: Real estate agents will be thrilled to have a personal assistant they can use to automate the toil in their work so they can focus on what matters: developing relationships and providing excellent service to their clients. Nobody wants to hunt through hundreds of menu items to accomplish a given task, which is why we are really excited by the success we are seeing with Lucy. I think that conversational interfaces will be the standard, rather than a “bonus” feature in the near future. Rechat is poised to be the front runner with such features in this industry.