SmithAI

The vast majority of the population won't open a command line or run a program to use AI. We simplify this process by letting anyone make custom AI Agents on demand without any technical experience.

Comment

Inspiration

We were inspired by an idea from Y Combinator that challenged us to make AI more accessible for everyone. Rather than limiting powerful AI tools to those who can navigate a terminal or write code, we saw an opportunity to democratize AI. Our project extends that mission by allowing anyone—even complete beginners—to harness AI for complex tasks on demand, opening up possibilities that were once reserved for highly technical users.

What It Does

SmithAI enables you to create and deploy AI “agents” tailored to your specific needs—be it answering questions, automating tasks, or analyzing data. Under the hood, SmithAI features an intelligent algorithm that determines which large language model (LLM) is best suited for your task and seamlessly routes your request there. This makes the process effortless for users: no more second-guessing which AI tool is the right fit, as SmithAI handles the complexity behind the scenes.

How We Built It

Full-Stack: We developed a frontend using HTML, JavaScript, and Tailwind CSS, and integrated it with ChatGPT, Ollama, Deepseek, and Claude on the backend. Smart Routing Algorithm: We created a core logic that evaluates multiple LLMs and dynamically chooses which one to use for a given query. App Store & Hosting: We set up our own app store on MongoDB and host each app on AWS EC2, utilizing Elastic Beanstalk for orchestration and S3 for storage. Scalable AI: We also run smaller AI models directly on EC2 instances, balancing performance with cost-effectiveness.

Challenges We Ran Into

One of our biggest hurdles was orchestrating multiple AI models with varying strengths and limitations. Ensuring each model could be invoked reliably—without bottlenecks or conflicts—required careful infrastructure planning and robust fallback mechanisms. Another challenge was building a user interface intuitive enough for non-technical users while still providing advanced features for power users.

Accomplishments That We're Proud Of

On-Demand AI Agents: We successfully built a system that allows anyone to spin up AI-driven agents without writing a single line of code. Seamless Model Switching: Our intelligent routing algorithm is a testament to our team’s engineering prowess—smoothly switching between different AI models to deliver the best results. Infrastructure Mastery: We’re thrilled with how we utilized AWS EC2, Elastic Beanstalk, and S3 in a cohesive way that scales as user demand grows.

What We Learned

We learned that integrating multiple AI models is as much about orchestration and engineering as it is about algorithmic intelligence. We discovered how to leverage the strengths of each model—like ChatGPT for natural language understanding or Claude for complex reasoning—and fuse them into a unified experience. We also honed our skills in full-stack development, AWS infrastructure management, and dynamic model selection based on performance metrics.

What's Next for SmithAI

We plan to:

  1. Expand Model Library: Incorporate even more specialized AI models for tasks like code generation, image recognition, and data analytics.
  2. User-Friendly Interface: Continue refining our UI/UX so that anyone can deploy advanced AI without feeling overwhelmed.
  3. Enhanced Collaboration: Introduce collaborative features that allow teams to build and share AI agents seamlessly.
  4. Global Accessibility: Pursue partnerships and multi-language support so our platform can serve diverse communities around the world. By lowering the barrier to entry for AI, we believe SmithAI can spark a new wave of innovation across industries and empower people of all backgrounds. The possibilities are limitless, and we’re excited to see how this technology will shape the future.
Share this project:

Updates