Jimmy's blog
photography, robots, drawings, and other "art"

Office layouts

June 12th, 2020


The first office I got to lay out was a robotics lab in grad school. We had a sandbox, a climbing wall, and a tree for robots to climb. One wall was entirely for tiny drawers with electrical and mechanical components. It was also right next to the machine shop. Oh yeah, there were some desks for humans to sit at. The sketch above took maybe 30 minutes of work and helped very quickly get everyone on the same page. I’ve found it’s worth drawing something out very detailed. Just “talking about it” gives room for miscommunication and will end up taking a lot more of your precious time and energy.

Office layouts are always in flux as the company grows or shrinks in size, moves buildings, or re-orgs. It affects everyone. I’ve been involved in more than a handful now. Some folk don’t care much where they sit, some folk get very anxious and want to chime in on everything. Add to those requirements driven by a machine shop, biology lab, kitchen, etc, it ends up causing a lot of hubbub. There’s a little bit of science behind office layout but mostly there are a lot of opinions. Given that there are a lot of subjective opinions and variables, here are some of the main design principles I’ve found and my subjective opinions sprinkled in :). Mostly for future me, but maybe you have some use for them.

Open office This is the most economical and least amount of effort design that we’re all familiar with. Unfortunately too often the executive team chooses this at the expense of productivity.
Caves and commons Private spaces (caves) where one can focus and work without being interrupted. This probably means you have your own room or cubicle. And close-by probably in a central location, a common area with shared desks and tables where team members can pull together for critical face-to-face time. This has been my favorite so far. It takes more space but I believe the increase in productivity is worth the extra cost.
Eudaimonia machine A set of rooms that by design you have to enter successively. They are arranged to optimize for maximum facetime and chatter in the first room to group work to increasingly quieter and private space for individual deep work in the last room. I’ve always wanted to try this out. It sounds like the most ideal workplace to me! Maybe one day.

Cross tribe mentality

June 5th, 2020

Humans naturally form tribes. I reckon it’s encoded in our lizard brain algorithms. Whether it’s seeking comfort, security, or status we naturally entrench in our narrow group and distrust outsiders. I consider myself lucky to have lived in different parts of the world to observe different tribes, but have never fancied myself as fully part of one. For better or worse, I seemingly feel like I don’t belong to a particular tribe, and the tribe seeking algorithm may not be as strong with me.

In friendships, I naturally gravitate to folk that are different from me rather than the same. I like to explore diverging viewpoints rather than finding the conventional common ground. I find myself looking at a group as an admiring outsider, become good at the virtues that the group holds important, and become accepted within the new tribe. Then as soon as I’m comfortable I become unhappy with myself to have become so entrenched and narrow minded that I’ve allowed myself to be surrounded only with like-minded folk before I eject myself and repeat this painful cycle all over again.

We are becoming more and more divided as a species right at a time when we are facing challenges that require global cooperation. Please distrust your inner tribe seeking algorithm, reach across to a different group, and challenge the norms and assumptions that your group holds. It may be a little painful but it also adds more color to your life.

Further reading:
[1] Third Culture Kids: for kids growing up in multiple cultures
[2] Cross Team Collaboration Cheat Sheet: for startups when you get beyond 20 employees and find yourself in the zone of multiple teams
[3] Team of Teams: what we can learn from empowering cross disciplinary local teams in the war in Afghanistan
[4] The Problem of Imagining the Real: anthropological view on our instincts on seeing

How to leave your company

May 25th, 2020

as a good leader and set the remaining team up for success

Research has shown that the way you recall memory is 50% determined by the peak intense point and 50% is determined by the ending [1]. So you might as well get the ending of your tenure at a company right. And let folk remember you as the classy person that you are. I’ve always believed the standard two weeks notice is okay for an individual contributor that has left clear documentation of their work but as a leader you have to start this process a lot earlier. Here’s what I see as a good timeline:

Four months before: have a successor in mind and start grooming this person. Regardless, you should always have trained your team to run without you in case you get hit by a bus.

Two months before: discuss seriously with your boss your own growth potential or whatever the reasons are that you are looking around and be transparent that you are exploring a bit if you have that kind of a relationship. Big if, I know. 

One month before: In the following successive order: tell your manager, discuss with the management team, discuss with your team, announce to the whole company. Ask them not to leak beyond the cluster of people you’re talking to.

Two weeks before: give the official official notice in writing (email) to your boss.

Last week: Give one last lecture. This doesn’t get written about much but I think it might be the most important step. Here you can reiterate why you think it’s important this company exists and that you still believe and why you want the company to succeed. But that for you personally, it was just the right time to pass the torch. In mine, I described the history of the company, tell the new employees some fun stories, lessons learned, give kudos to your successor, and anything else you want the company to know. Leave on a positive note!

Last day: Say your goodbyes but be respectful and don’t be a distraction the whole day long.

One month after: Write a nice blog post on social media on why you left. Keep it positive.  Here’s mine.

Two months after: This is when you can announce your new job. I like to give it some space out of respect for the company you left and I also like to see how things work out at the new company.


If you’re able to time it, leave on a high note when the company is doing well, rather than when there’s a crisis which is when most mice leave the ship. Though at the time one can’t tell whether they are lesser beings or lesser fools.

Edit: I wrote this blog post right after I left Strateos beginning of 2019, before covid struck mid March of 2020. So that last sentence is rather ironic.

Running for the cure

April 27th, 2020

I’ve run a half marathon before. During this shelter in place directive, it seems like now is as good a time as any to go for that full marathon. You can follow my progress on Strava:

strava heat map

I’m thinking to run a half marathon this quarter and aim for a full marathon by end of Q3 and establish a route through the trails in Pacifica. Unfortunately, previous attempts by runners to find a cure for a disease have not been found to be fruitful yet. At a race in Atlanta for example: “despite their diligent, dedicated running, the 6,000-plus participants in a 5K Race For The Cure did not find a cure for breast cancer”. Original report here.

A machine learning based clock

January 12th, 2020

I wanted to gain more understanding of machine learning. I learn best by doing and also find that I can fool myself (and others) into thinking I know something by reading a book and watching a couple videos. But until I implement something myself, I don’t really know it. As Richard Feynman said: “what I cannot create, I do not understand.” So I built a clock that uses a neural net to tell time by looking at the sky.

I made a web app here so you can try it yourself.

My full dataset can be found here. These are 30 second time-lapses from midnight to midnight at 30fps. So each movie has 900 frames taken over 24 hours. I have 28 days worth of data and counting. Below is the full confusion matrix on my training data. The classifier doesn’t work so well when it’s fully cloudy or foggy otherwise I was surprised to learn it can be +/- 1 hour on a clear day and totally able to distinguish between morning and evening. So it must look at features other than just brightness.


Everything here I learnt from Jeremy Howard’s excellent fast.ai class.

Next chapter for Jimmy and Transcriptic/Strateos

December 29th, 2019

We started Strateos (formerly Transcriptic) as outsiders to the industry with little experience but enough chutzpah (aka ignorance) to think we could make biology easier to engineer. We charged a whopping $100 for our first run of growth curves to a class of undergraduates at Stanford. I was enamored by the science being executed at our facility and the idea that somewhere remote, college students were designing how to grow bacteria from their laptops in their dorm rooms. We realized that to keep doing cool science we had to prove that we could turn this into an actual business that is larger than $100 checks. And so it started.

Today, we work with the biggest pharma and synthetic biology companies in the world. We went from a couple of us manually pipetting to today where we have dozens of robot arms managing hundreds of instruments. We are deployed across three different sites across the US and just surpassed a million unit operations in our Menlo Park facility alone. We have also recently expanded into automated chemistry to rapidly synthesize new compounds that we can use to probe the molecular biology we have been running. It’s safe to say that the checks are a little bigger now and the question is not if there is a business but how to keep growing and keep up with the demand. As I mentioned, the company started off as audacious (and ignorant) outsiders but we have complemented the team with insiders and decades of experience by the likes of Mark Fischer-Colbrie (former CEO of Labcyte) and Dan Sipes (former Director of Automation at GNF). These were the very companies I studied when we were first starting out! I could not have imagined that those leaders would eventually end up joining our little company.

Like any startup, we’ve had our share of ups and downs along the journey. There was a short period of time where we were living through season 3, episode 5 of Silicon Valley: The Empty Chair by not having a CEO. I kid you not. Fortunately, it turns out that keeping the company going without a CEO actually wasn’t all that bad. Mergers are hard but the dust has settled now and we’ve survived that one too and come out stronger. It’s a funny thing: when the company is facing tough challenges, I can’t leave because I want to fix the situation we’re in and be the steady rock for the team. When the company is on a high, it’s so much fun that there’s no reason for me to leave. I’ve been riding those ups and downs for over six years now! Today, I believe the company is in the good hands of experienced leaders and is finally able to stand on its own legs with steady business incoming. I believe it is the right time for me to pass the torch.

Friends come and go. A day ends, a new day begins. What’s important is to make it meaningful: a meaningful friend, a meaningful day. For me, it is time to re-pot. To current and former Transcriptic/Stateos employees: what I will look back on is the hours spent alongside all of you, overcoming challenges and forging deep relationships along the way. If there’s a particularly meaningful memory you have of us, from the heroic to just something that made you laugh, send me a note! I’d love to hear from you. To potentially new Strateos employees: now is a super exciting time to join. We are about to turn on our largest data generating machine yet! More news on that to come in the new year. Stay tuned.


(This is a cross-post from LinkedIn. Added here for consistency and completeness.)

Reflecting on how much easier it is to take a panorama compared to 10 years ago

May 26th, 2019

2400 Chestnut pano3cropped
This photo was taken circa 2007 in Philadelphia, PA. Back then digital photography was still becoming popular and panoramas weren’t as easy as a rotating your phone. This is a combination of HDR (blending 3 photos with 3 different exposures taken from the same position) and panaroma (stitching together photos from 10 different positions) for a total of 30 images combined into one image. From setting up a tripod, pressing the shutter on my DSLR, stitching the images together using Autopano Pro software and tuning the HDR it was about 8 hours worth of work. An oldie but still one of my more favorite panoramas I’ve taken and symbolizing a previous chapter in my life. Thanks friends and city of Philadelphia for all the adventures and life lessons. Some of you I take with me on my journey, parts of you I sadly must leave behind.

To a new chapter in Pacifica, CA. My new skyline going forward. This image took about 5 seconds to make with a single click and sweeping across the landscape with my phone.
Pacifica Skyline

Robot sketches

September 13th, 2016

Ink and copic marker.

Landscape sketches

September 13th, 2016

Ink and copic marker.

Random Walkers

May 30th, 2016