Meet Anthony Mein, one of our pioneering Forward Deployed Engineers, and learn about his journey from biology lab AI to building enterprise AI agents at the forefront of customer experience.
Watch the full interview on Forward-Deployed Caffeine podcast, lead by Cresta’s Head of Forward Deployed Engineering, Jove Zhong.
----
Tell us about yourself - what is your background?
I'm a Forward Deployed Engineer (FDE) here at Cresta. I've been here for four months, and I was actually the third full-time FDE hired. Before Cresta, I was working on applications for reasoning models and agentic AI in the biology lab sciences space. It was an augmented reality (AR) plus AI application where I was one of two developers doing the entire stack—web platform, AR client, and AI tools.
The transition to Cresta has been amazing because I really love working with customers and seeing how customer-facing our product is. The AI agents we build interact with thousands of customers every day in production, and it's incredible to see the real-world impact.
What inspired you to join Cresta?
I was really looking to learn from more experienced developers and get into a space that's fast-moving with great engineers. When I arrived at Cresta, I wasn't exactly sure what to expect, but something that surprised me so much in my first month was just how fast the team moves.
The breakneck pace isn't stressful—everyone is just so excited to work at this edge of AI technology. It feels very similar to the quantum mechanics revolution. After the advent of LLMs, there's this huge rush to find best practices, improve the models, and figure out what works. There's so much room for innovation and growth.
What is your role at Cresta?
As a Forward Deployed Engineer, I work on client-specific implementations rather than generalization. I identify patterns among customer deployments that we can bring back to the platform and reuse. We act as that bridge between the customer and our product team.
Success isn't just driven by product metrics; our main driving metric is ensuring that the customers we're working with are successful. We want this AI agent deployment to be successful for them, and we need to guide them with deep AI expertise that they might not have in-house.
What does a typical day look like?
Day to day involves a lot of client interaction. I'll be building an agent and asking questions to the customer about how they want it to respond to certain scenarios. Your job is embedded in client success, so you form this great relationship because success matters as much to you as it does to them.
After weeks of building the agent for one of my first deployments, we launched to traffic. The customer said they couldn't believe how we fully broke down what the agent was doing and how quickly we fixed issues. I remember the first day releasing to production…I was so nervous, but everything went smoothly thanks to our robust testing framework.
Can you share more about the testing and validation work?
Simulated visitors are a huge aspect of testing AI agents. Because we have Agent Assist and Conversation Intelligence, we're uniquely positioned to make our simulated visitors much more realistic. We fine-tune smaller open-source models to act as simulated visitors grounded in real conversations, tuned specifically to the end customers.
If you just give an LLM a prompt like "you're an angry customer", it sounds so generic and verbose. Our approach creates agents that respond like real people would respond, testing thousands of conversations before hitting live traffic.
What makes Cresta's approach different from customers building their own?
In the call center space, customers would have to build AI agents that are not just a prompt and tools. You've got to worry about low latency, Automatic Speech Recognition (ASR), turn detection (how do we know when a customer stopped speaking)? That framework is a large undertaking.
At Cresta, we've been working on this space for years. We have the expertise and best framework for turn detection. We're at the front line of technology, ensuring that new practices are folded into the product. A customer would need 30-40 engineers for at least a year to build this framework, plus a dedicated team to keep up with latest practices.
That's our spot in the industry: we handle best practices so customers can focus on their bread and butter: customer interactions.
What about AI and human collaboration?
I really respect that Cresta is very interested in human and AI agents in unison. We want humans in the loop. I go to top call center leaders and top agents, and AI is a long way from replacing them. Humans are really good at solving intricate problems.
What our AI agents do is take so much workload off of them so they can be even more present for the customer. There's a really good synergy between human agents and AI agents. For repeatable workflows like replacing a credit card, customers don't care if it's AI. But for complex issues where they're frustrated or concerned, that's where human agents come in and really solve those issues.
What advice would you give to aspiring FDEs or students interested in AI?
As a whole, we're seeing the entire coding industry change overnight. You need to be familiar with AI tools, but the most important part is learning how to read code and understand what makes good code. Ground your work with AI in real knowledge.
I graduated college in 2022, so ChatGPT didn't really come onto the scene until 2023. I'm really thankful I had to learn things by hand because now when I'm working with AI, most of my job feels like I'm spending more time reading code than writing code.
For FDEs specifically, you need to get your people skills up and be able to have friendly interactions with customers. You have to know when to say no to a client, manage expectations, and be friendly about it. Every client sometimes wants Artificial General Intelligence (AGI), and you can't promise them AGI.
What gets you excited about working at Cresta?
The company's direction for AI is driven by FDEs. We interact with customers every day, understand their wishes, and bring feedback back to the platform. When I tell a customer "let me go talk to our platform team", and we come back saying "we're going to build this for you", it's a great interaction. They feel—and they’re right—that we're really personally invested in their success alongside ours.
As we solve more problems and make AI agents more adaptable, we snowball in how complex of deployments we can handle. It's a pleasure to work on really interesting stuff. The path is clear: build bespoke solutions for customers, and the positive iteration loop creates solutions they really notice.
Curious about Forward Deployed Engineering or building enterprise AI agents? Cresta is hiring FDEs across different domains. Check out our careers page to learn more.
And for an in-depth look at evaluating, selecting, and deploying AI agents with confidence, download The Enterprise Buyer's Guide to AI Agents.
