My Website
Charlie Angela Mehlenbeck

This site is under construction with very little time on my hands.

Please reach out to me on my LinkedIn for now and check back soon for more content.

A picture of me

Take a look at some pretty pictures here too or feel free to look at my CV here until I finish being picky with my js.

For those who know me already and haven't seen or heard from me in a while:

After I left Dematic, I spent some time learning FPGA's and built a robot with a friend that we were going to compete with. Techy stuff: It had mechanum drive with a custom motor driver board connected to a Xilinx Zynq FPGA (SOC) for control running a custom linux rootfs (built with petalinux) with custom AXI interfaces between the FPGA fabric and the arm, with custom linux drivers, talking over ethernet to an onboard Intel SBC running ROS with remote control over WIFI and 3D SLAM functioning. I was even in the process of equipping it with a thermal camera through the FPGA fabric (probably be much easier now even with ChatGPT's help with my RTL, having tested it's coding capability extensively, including VHDL). I was even integrating high speed motor current sensing which I was going to use to make predictive traction control using in hardware NNs.

Some personal drama happened there but, shortly after I came up with an idea for my first company and then covid hit. During that time, I played the stock market and did pretty well for my family and I, more than enough to try starting my company. (It's mildly amusing to me how people sold the companies making the computers and IC's that facilitated vaccine development and drug discovery for a then unknown disease (and all others and more) and then bought companies they didn't know if would succeed in devising a cure or vaccine.)

A different friend of mine however then lost his job to covid and I tried starting the company with him. We started with a novel soft robotics actuator. I did TONS of research into things like lost wax casting of soft parts, silicones, fluid based computation, and microfluidics. I thought I got something that could produce what I'm after figured out, and we even built out a lot of the production process in my barn. We started to test it and even started on some custom software to model it programatically so we could vary the design, but it would have meant too much further product development afterwords to make a viable product in time. I'm pretty confident I have manufacturing figured out in theory, but development and testing of it as a production method (even at small scale) would have been too much to chew on for a first go, so we 'quickly' pivoted to Agricultural Robotics thinking it would be easier with the amount of open source work we could draw on.

The new goal was "Seed Weed and Water the home garden cheaply and efficiently at scale". It was going pretty good for about a year and after a lot of mentoring and practice working together on the actuator, we were working very well together. We were like family, and each found a new extended family. We very nearly had a viable prototype of a rather complex but very cheep yet accurate and scalable swarm robotic system along with everything to test it, simulate it, and even use it to farm a small acreage. We had 3d scanning in the barn to validate at cost localization in GPS denied locations. We had servers set up with our own intranet for training, in-house development, and local large scale data collection (too much for any remotely accessible internet at the time). We had maximal use of FOSS S & H in all levels of development. We had the imaging pipeline to be able to train segmentation models ready and applied to various tasks. We had 4 different robot platforms built, 2 entirely from scratch with totally custom localization, path planning, sensor fusion, time keeping, networking, and PCBs using cheep MCUs. It was all written in a way that was platform and MCU agnostic so we would be essentially immune to the chip shortages that were then becoming so common. We even had a 3D simulation in Unity that would let us validate the firmware with maximal code reuse.

In the end however, my co-founder had a friend going through a really bad time and he and his partner took them in with a warm heart but very little experience in their needs which were much greater than he expected. I stayed to help (us all living together on off) and it consumed voluntarily (and not regrettably) all of our time for almost a year. Unfortunately, while we all 4 of us in that massively chaotic time learned a absolutely huge amount about humanity that I would never dare regret, and I met the first person I've dated in my life, it took a tole on all of us and it virtually destroyed his and my friendship to the point where we're each now having to go our separate ways with virtually nothing left in hand.

Shortly after loosing relative touch with my then co-founder, I started working with ChatGPT as it became popular and then learning about LLMs in general. Having never liked LSTMs at the time I had then last played with anything AI, given their (IMHO) pretty losy in-elegant memory scheme, I've been mainly testing transformer models in cloud compute and then looking to see if I can integrate them directly with the linux terminal effectively so they can be re-trained to work on much larger programming tasks. A few people have been poking at this beside me, I'm just one of them at this point (my work on that is GPL on github), but... occasionally I dabble on that with the hope I might soon be able to talk virbally with a personal assistant that can code for me on a hike or in the car so we can all as engineers (and people in general!) gain a bit more humanity and a bit less screen time with full *efficient* audio control of an environment.

I occasionally pluck at what's left of the code base we left, but at this point I'm again looking for work in robotics, aerospace, CV, ML, anything interesting. While of course, spending a lot of time with my partner who has started door dashing with me in our spare time as we take care of a new puppy, several cats, a messy house and an apartment, and some stuff I probably shouldn't talk about or I'll make them anxious. :P

Very fortunately however, and fairly expectedly & understandably, and ??maybe?? not, however... There's not much time for any of it atm. :)

If you're curious though, you can see our company site here at Mind the speed of the site though, it loads slow. We were just getting our stuff together to start meeting investors and prospective customers, so... our home brew wordpress site was on a bit small of a instance for all the themes. (Now I code this site in straight php, but, we tried.)