About dragonflys.ai
Our Mission
- At Dragonflys.ai, we are passionate about redefining the future of AI computation in hardware.
- Our mission is to create massively parallel FPGA Systems On a Chip (SOC).
- Our technology accelerates inference processes in embedded systems such as drones or space flight computers enabling AI technologies to fully utilize full hardware acceleration.
- We are bringing forth the next generation of efficient, reliable, and powerful AI platforms through developing a state-of-the-art interpreter/compiler engine, along with a more intelligent AI based OS integrated directly in hardware.
- Imagine programmable hardware acceleration tuned for your specific application.
Our new CORE I FPGA AI Playground Development Board is shown below....
Our Technology
Thinking outside the box
- Computer science has always promoted the concept of writing software using a fixed set of opcodes designed by someone else. Central Processing Units (CPU) chips today are designed by a committee for general purpose computing: one size fits all.
- Using Field Programable Gate Arrays (FPGAs) hardware in place of CPUs, you can now create your own complex opcodes in order to accelerate your applications. With our technology, your software calls them just like any other software function in your code. We call these MPUs (Massive Processing Units), opcodes and functions are called "Words" to our AI machine. They operate at up to 250mhz each process. You can have thousands of processes on a single FPGA chip.
- Words are the basis for truly Natural Language Processing (NLP).
- You can even define your own new Words quickly and interactively in the terminal, using the built-in interpreter/incremental compiler.
- Thousands of concurrent processes can run in parallel on a single FPGA chip. These parallel processing units can be controlled and accessed by either new Words you define of opcodes you create yourself.
- The Built-in OS, Interpreter and Compiler literally creates a full AI capable computer on a chip.
- The Chatbot User Interface allows you to talk to the chip just like any other AI chatbot: Bard, Claude or ChatGPT.
Value proposition of the company
- Dragonflys.ai provides a unique combination of both hardware and software integrated solutions leading the way to more powerful and easier to program/maintain AI On a Chip (AOC).
- Innovation in Integration: Our interpreter and compiler engine plus the OS are designed specifically for FPGA chips, which delivers unparalleled performance and efficiency.
- Speed and Agility: Our FPGA based technology is optimized for faster training and inference, drastically reducing the time it takes to develop and deploy AI models.
- Customized Solutions: We cater to diverse needs and applications, providing customizable chip based solutions for our industrial partners.
- Affordability & Scalability: Your system can be implemented in one or more FPGAs. Use as an embedded AI system or network dozens of our FPGAs together in order to build a massively parallel AI server.
Applications: AI, robotics, telecommunications, space
AI and Machine Learning
AI,LLM, ML game-changer Technology...
Dragonflys.ai FPGA chip technology is a complete game-changer in fields of AI and machine learning. The system supports the use of massive parallel processing providing the speed required to train complex models, analyze large datasets, plus make real-time decisions directly in hardware.
This illustration depicts NASA’s Juno spacecraft soaring over Jupiter’s south pole.
Credits: NASA/JPL-Caltech
Robotics
Agile and intelligent robots...
Our FPGA based technology enables more agile and intelligent robots operating in real-time. By speeding up the training process, robots can learn and adapt more quickly, resulting in smarter automation.
The Triune OS/Language using behavior based coding, gives the robot the ability to make intelligent decisions in real-time.
Using our Triune OS and language modeled on the human brain (Triune Brain), behavioral based intelligent control system gives you the ability to teach the robot like teaching a child - the robot can also teach itself.
Space
More intelligent, remote programming, lower cost space avionics...
Using space rated FPGAs with our technology, space based systems can now be reprogrammed without shutting down the robotic spacecraft for programming. No need for complex Linux based systems. Interactively communicate with the spacecraft over low speed data links. Upload source code - it compiles within the spacecraft in real-time. In just a few clock cycles, this new code is incorporated into the system, seamlessly without loss of control of the spacecraft for even a second.
If you had one hundred space robots building a structure in deep space, how can the main software be updated over a three hundred byte per second data link today? Currently, you have to transmit a large multi-megabyte binary blob then JTAG the chip. With dragonflys.ai technology, you just send up a couple of hundred characters of text to update the software and the spacecraft doesn't miss a beat.