Employee interview: Andreas Dehnhardt, VP Applications Engineering, Co-founder
This is the second in a series of employee interviews. Today we speak to Andreas, who’s been leading the software engineering department at videantis since the company spun out in 2004.
2004 is a while ago; which accomplishment in those 10 years are you most proud of?
I could tell you about all the software we’ve ported and optimized, or the clever new architectural features we’ve developed, but truth is that it’s much simpler than that. What I’m most proud of is that everything we’ve delivered to our customers worked as promised and is competitive. We have many customers that are shipping products in volume, and there’s nothing more satisfying than being able to buy a product that you know has your technology inside. Whether it is a car in a showroom, or a piece of electronics at the local Media Markt or Fry’s, I always think: “my stuff is in there”, and that’s something to be proud of. To get to market, and into volume, our customers have to beat their strongest competitors, which isn’t easy. So getting into volume, and getting repeat business, is also proof that it was easy to integrate our product, and that we helped our customers go to market on time, at cost, and with a competitive set of features.
What does a typical day look like for you?
I always come into the office by bike, no matter what the weather. First thing I do is grab a coffee, then start up my laptop, like most people do I guess. Next is to type up the ideas I had overnight or under the shower in the morning. It’s interesting how the brain works – during down time, these creative ideas tend to bubble up. Oftentimes these are ideas on how to solve different software optimization problems.
Then, it’s back to business. A significant chunk of the day usually consists of conference calls or face-to-face meetings with customers or internal meetings with colleagues. On the development side, I function as a key interface between the different teams, for example our algorithms team and optimization team. For the video coding and computer vision applications we work on, it’s an absolute necessity to understand the algorithms in order to build the most efficient implementations. Another key task for me, which is actually more of a continuous process, is to combine the feedback from our target markets and optimization strategies into recommendations on how to improve our hardware architecture. For that, I work closely with our microarchitecture and processor designers. For the software development tools, the process is similar. The key purpose of tools is to help smoothly port, optimize, and verify algorithms and applications, so I am intimately involved with steering the direction there too. We have very short lines within the company, so it happens very often that issues that arise are addressed the same day, which is always gratifying.
What do you see as the biggest opportunities for videantis and its customers?
When I look at the processors that are out there, I believe none of them really address what the market needs. We still have customers from projects that we did some time ago calling us and telling us that we really have something unique. Scaling up in performance is hard for instance. We’ve seen quite a few companies that tried to scale up their programmable video processing engines from SD to HD resolutions, and failed. We tackled this scaling problem quite some time ago, and taped out our 10-core test chip already in 2010. We’ve been optimizing all our software for multicore for a long time, both our codecs as well as our vision algorithms. Another problem is performance. We see most companies addressing this with wide SIMD architectures. This works for a small class of algorithms, but doesn’t work well for a wider variety of algorithms that require smaller block-based processing or control at the pixel level. They’re adding contrived architectural solutions, or are hard-wiring those algorithms that don’t map well. Our architecture combines performance, low power, small area and real flexibility in a single unique package, allowing us to run both video codecs as well as computer vision algorithms on the same scalable architecture. Embedded vision, video and image processing are hot topics, so addressing this with our technology is a big opportunity.
What are some key challenges you’re seeing?
One challenge is, as always perhaps, to find the right talent. As we’re scaling up the company, it’s a challenge to find people that have a good mix of hardware, software and video/vision experience. Good people attract good people though, so that helps.
Another key challenge is automotive safety. Our technology is integrated into ADAS systems, which greatly improves automotive safety, but it’s actually quite hard to build a system that’s 100% safe. We see people often underestimate the workload. Things that work in consumer electronics require a lot more effort to do right in automotive. When your TV’s set top box reboots once in a while for instance, or stutters a bit, that’s not a huge issue. In a car, lives are at stake. This has big implications on the hardware and software design. Luckily, we have some unique technology that addresses automotive safety.
Coming back to scaling up the company. You have several open positions for your team, can you tell us what you’re looking for in candidates?
The ideal candidate is someone who likes to work in an environment that is focused on getting clever ideas into real products at our customers. A combination of having experience with computer vision, video codecs, and DSP processor architectures is ideal; people who understand software, algorithms, processors, and hardware. Most computer science programs teach programming using higher-level languages these days, like Java, with little focus on the hardware underneath. Since we do a lot of platform-specific optimizations, we look for people that like to understand what is actually running the software too.
What do you like most in your job?
Even though I’m a technology guy, I like working with my colleagues and customers most. The team has been really stable, and the founders are still all with the company, which is pretty unique for a 10-year old company. It’s nice to grow the company together. Of course it’s exciting to push the envelope of video processing too.
Anything you’d like to say to our readers?
It’s great to be working on technology that can really change the world. We’re working on automotive technology that saves lives. Bringing embedded vision to consumer electronics like mobile phones and tablets enables new user experiences. Longer term, I believe embedded vision will play a key role in wearables, health and internet-of-things applications. It’s exciting to be here. Join us on this journey!