Hi, I'm Greg, and I like to build things.
I'm writing this in July 2024. At the moment, I primarily work with LLMs - OpenAI's ChatGPT, Anthropic's Claude, and open source models like Meta's Llama 3 and Google's Gemma 2.
I started working with LLMs shortly after the full version of ChatGPT 2 was released in late 2019, and I haven't stopped. My background is in product management, but over the past few years I've accidentally become more of an AI 'integration specialist' - I don't build LLMs from scratch, but I tune, tweak, and twist them to make them solve problems better than anything off-the-shelf.
I'm currently the sole full-time technologist at a startup called Symphony42. We build persuasive AI, and we use it to help our clients sell things. It works pretty well, and developing it is incredibly fun.
I also work on a small number of side projects.
If there's one thing I believe, it's this:
The next decade is going to be extremely weird.
I don't believe this because I think the Singularity is around the corner, or that we're necessarily going to create artificial superintelligence. I believe things are about to get weird because the technologicial advancements we've made already, fully deployed and distributed, are enough on their own to fundamentally transform society. Any additional progress beyond that is just upside.
This is going to be amazing!
Yes, there's going to be some disruption. If your job is boring and routine and done over the phone or through a computer, it's not going to be your job for much longer. Not everyone is going to be a winner in the age of prevalent AI. But if you're willing to grab the opportunity - the tools out there are getting better and better, and that grabbing's only getting easier. We can all do so much more with less. Things we were previously incapable of doing are now within reach.
Artificial intelligence isn't going to replace us; it's going to augment us.
So, here's what I can do.
- Build with the APIs and SDKs of the leading AI providers, including OpenAI and Anthropic
- Design and implement AI-powered automation for customer service channels
- Deploy and scale open source language models on cloud platforms, including AWS and Azure
- Create lifelike digital avatars using voice cloning and generative video
- Fine-tune open source language models for domain-specific tasks and improved performance
- Implement efficient prompt engineering and chaining strategies with LangChain for advanced reasoning tasks
- Use PyTorch and HuggingFace's Transformers library for rapid prototyping and deployment
- Orchestrate multi-agent AI systems using LangGraph for coordinated task completion
- Develop conversational AI systems using text-to-speech, SMS, and telephony services
- Produce compelling interfaces and APIs for the products I create
There has never been a better time to get building.
Before I became aware of LLMs, I thought I was permanently retired. I was smart lucky enough to merge my last company into another one that did ridiculously well, and I just didn't have to work.
But the new AI-powered technology boom was simply too compelling to resist. I couldn't stay out - it was clearly time to build.
You can feel it too, can't you? If not, surely you would've stopped reading by now. I'm an irredeemable techno-optimist, and writing with enthusiasm, optimism, and excitement about the future is kryptonite to non-builders. Their eyes roll, their lips curl, and they move on. But you're still here.