My name is Ankur Mohan. I’m a Technology Architect in AI and Machine Learning and the author of this blog. I love technology and the power of technology to change life for the better and make people more productive. I’m originally from India and got my Bachelors in Electrical Engineering from IIT-Bombay, followed by a M.S. in Electrical Engineering from the University of Maryland and a M.B.A. from Georgetown University.
I did computer vision research during grad school and in my first couple of jobs after grad school. Following that, I was an early member of a startup called Scaleform that developed a flash player runtime for video games. This startup became successful and was acquired by Autodesk, where I was a product manager for many years. Around 2015, I became interested in drones and flight control systems – I built a drone from scratch, including writing the flight control system (sensor fusion, command and PID control system, PWM mixer and many other modules). This was by far the hardest thing I have done 🙂 A video of my drone’s first flight is here:
Nothing super exciting, but the Arduino Mega board on the quadcopter is running my flight control system to stay level! It is getting raw sensor inputs from the gyros and accels, computing orientation, sending the right control signals to the motors, accepting user input (I was controlling the throttle while recording the video, which is why the drone is not always in view in the video)..
Another one I recorded while I was adjusting the PID control gains using “one-axis control”:
You can also see the command and control application in this video.
Seeing the huge potential of drones in capturing aerial imagery for surveys, I worked with a friend at a DC area architecture firm and did some drone based land surveys. Some of the results are shown in these pages (which were built to showcase the potential of drone based aerial imagery to construct geotiffs and point clouds). The sidebar shows WebGL views of 3D models constructed by applying photogrammetry to drone imagery. You can use the mouse to rotate these models in 3D.
Over the last three year, I have been focusing on AI and I think I have found the perfect field to spend the rest of my career. AI is a great blend of mathematics and statistics, software engineering and hardware engineering. The confluence of availability of large amounts of data and advances in neural network architecture is resulting in almost magical-seeming solutions to problems such as object localization and recognition which have long been considered very difficult to solve by computers. This means that AI has paradigm changing applications in every industry and will touch all of our lives. I have read hundreds of AI related papers over the last year and worked on some significant projects. I have benefited tremendously from articles and blogs written by other people, and hope that my blogs will be useful for others. Currently, AI research is open and collaborative. Researchers release source code for their work, making it possible for others to learn and reproduce results. People like me read the papers and dig through the code and write blogs that hopefully make some of the finer points easier to understand for the community. I hope this spirit of openness, sharing and collaboration continues!
I have noticed that among non-technical people, there can be significant misunderstanding about what AI is and what it is currently capable of doing. Sometimes, this misunderstanding borders on fear and misapprehension. After all, you fear what you do not understand. During my conversations with my non-technical friends, I attempt to (so far through 1-1 conversations, but hopefully in the future, in a more organized way) help them understand what AI is and how it works at a basic level in simple, non-technical language.