Worked for a defense contractor

Got fired from Technovative Applications, at just shy of seven years. How I got there:

Technovative Applications is a small defense contractor started by Jim Williams in the early 1990s. Liked Jim on first meeting. Was sure he was a damned good radar Engineer. Jim started with a different notion of how to build radars. Believed he was likely right.

As back-story, Jim's company (TA) mostly built specialized radars as test articles for various military projects. The test-ranges started to refer to TA purpose-built radars as "truth sensors", meaning they assumed Jim's radars were right, and everyone else less.

This outfit is a last gasp of the Cold War. Jim had collected a bunch of old cold-warrior-engineers to help out, who are clearly brilliant. But ... Jim is also an older guy - in his 80s. As such, he tends to forget things. Much of his old crew has died. Still, some bright guys, but also very much a limited time chance.

A research arm of the DoD finally gave TA money to build a bigger radar. That first big radar was called PFCR (Precision Fire Control Radar). I came in when the physical radar was assembled, but not working.

(For any security geeks tuning in, will only describe what is already in prior DoD missives, and no other specifics.)

Looked around, and suspected the Linux driver and FPGA firmware was the greatest risk. (The other software folk knew the Linux driver was a mess, and did not want to touch.)

Nearly lost my mind for about five months.

Turned out the radar hardware got squirrely within one specific window. The radar would crash - hard - in the middle of a mission (when tracking). In the Linux driver, precluded any access to the hardware interface in that window, and the radar became solidly reliable. Required some very precise calculation of time ... but worked!

This was very much a near thing. I did not make the radar brilliant. I made a brilliant radar reliable.

Turns out, Jim forgot this.

That radar went on to do remarkable things (and still), and prove dead reliable.

Keep in mind, something like Patriot anti-missile system is expensive. Patriot puts a radar and computer in the missile - and this is very(!) expensive. If your missile interceptor is more expensive than the missile you intercept, the economics are all wrong.

On the other hand, if you have a sufficiently precise ground-based radar, you can take the radar (and computer) out of the interceptor. When the interceptor is much simpler / less expensive than the threat, a single cargo plane-load can stop a lot.

The success of that first radar lead to the DoD funding several larger radars. This is where things went sideways.

Got to see the development of next batch of radars, from the start. Jim's crew was used to throwing together simpler radars, and hacking through any problems found. With simpler radars, and capable folk, this approach works. With more complex radars - not so much.

The program slipped, badly. But the DoD had nothing as good, so held on. Even funded a follow-on program.

Worked on that next radar, from the start. Designed a much more powerful radar computer, and got buy-in from the military. Have worked aggressive development at small companies and startups most of my working life. Knew how to manage risk. Talked to just about everyone in Engineering. Thought I had a good chance to keep risk in check, and make the timeline.

Was trying to teach an old dog new tricks. Guess that did not work. :/