Why Trump Sucks

uid: 202010071300 tags: #politics

February 22, 2023

Reflecting On A Negative Teaching Experience

Find and describe an example of a negative experience/interaction with a course staff member. Include a reflection on what specific actions, assumptions, and/or factors made the experience/interaction negative and how it could’ve been handled better. Be prepared to talk about your examples in section next week.

When I was a student in CS61C in Spring 2019, the workload for the class was ridiculously high - the most workload I’ve had out of any class I’ve taken at Berkeley. In addition, there was a large discrepancy between the current offering of the class and how it was structured when Professor Garcia taught it — Garcia only had 3 projects instead of 5 long ones, fewer homeworks, fewer projects, and a significantly more generous grading curve. At one point late in the semester, tensions flared up between students and course staff — a student made a long Piazza post detailing all of their complaints about the course, and the head TA responded by essentially stating that CS61C was a high workload class by default, and while the course staff would try to change the structure of the projects and grading schedule to make things easier for students in the future, there wasn’t anything significant they could do about it currently. (This was a couple years ago, so my memory is a little fuzzy, but I think this was the general gist.) This caused more tensions, because I think the head TA had a different idea about the amount of time the projects were taking than the students, and there was a significant gap between what the TAs thought the students would be a able to handle and what the average student was actually capable of handling. Another thing that I think made the interaction worse was the tone of the TA - it was fairly removed and aloof, trying to reason about the workload of the class objectively while the issue had devolved into a very emotional matter at that point.

Eventually, the situation was resolved by the Professor himself, Professor Weaver, who committed to honoring the same grading curve that Professor Garcia had, while also promising that he would take all of the complaints and concerns to heart and make the class better for future iterations of the course. It was a genuine and heartfelt response, and it seemed to cool down tensions quite a bit. Because of the way Professor handled the matter in the end, I think most students were ultimately happy with how the class (and their grade) turned out.

uid: 202010020937 tags: #teaching

February 22, 2023

Human Capital Stealth Networking Event (_)

Agora (Ryan Gibson)

San Francisco, https://www.helloagora.com/

Construction industry - 10T/year, employing over 200 million a year
- Productivity has remained flat over the last 20 years (big opportunity)
Construction materials operation is broken

Challenges

Product: A. Building 0 to 1 tools for a complex legacy industry B. Transformation of the industry Looking for focused, product- and customer- driven engineers

Kanishk is working on revamping underlying data with new ETL, Faceted Search, Relevancy, Custom Items Infra: ElasticSearch revamp

Vise (Nikki Kang)

New York, https://vise.com/

Want to become the central nervous system of the modern financial advisor’s practice.
Platform for portfolio management, that financial advisors can use, so they can focus on client relationships

Technical challenges

Building the portfolio construction engine
Data visualizations; organization and presentation of dense information
Integrating with custodians and other external data
Foundational architecture across the stack

Q/A

New grad recruitment
- Recruiting lead
- 1-2 phone screens
- Onsite
Role of the financial/quant engineer
- Quant engineers are working on RPC (portfolio construction engine)
How did your role scale as the engineering team grew?
- Lots of opportunities, product design, recruiting, etc
What does a typical day look like?
- Every day standup, sometimes there’s a few meetings about product-decisions, drawing up tech specs with the eng teams
- Every week, there’s a sprint
Team dynamic
- Everyone’s celebrating being back in the office, working side by side
- Everyone on the team is great, excited about the way the team is growing
Are they hiring applied AI engineers
- Yes
Greatest learning experience
- Growing a startup, things are a lot faster compared to Airbnb, have to be more careful/accountable with what you’re shipping
Do Vise employees abide by FINRA regulations
- No
New grad, mentorship
- Great team, learn a lot from them (not a formal mentorship program), everyone’s willing to teach people things
What responsibilities can you expect as a new grad
- Depends on what the role is, parts of the product that people are trying to fix ASAP

Observe (Jacob)

San Mateo, https://www.observeinc.com/, Jacob Leverich jacob@observeinc.com

Want to become market leader in observability
One platform for all of a company’s machine data
Frontend (Typescript/React), Backend (Go)
Heavy management is done in Snowflake

Q/A

Still have to work on alerts
32 people
Still early days, reaching out to people

Qualia (Joel Gottsegen)

Qualia, San Francisco (https://www.qualia.com/)

100M raised, 200-300 employees
Make it easy to buy and sell homes
Core software - Workflow software for title companies, mortgage lenders (backend of real estate transaction)
There is also some front-end software, used for buyers and sellers of a house
Use Meteor
Pretty huge industry

Q/A

Qualia open-source stuff https://github.com/qualialabs/reval https://github.com/qualialabs/web-shell
Front-end — semantic UI
Also pretty much own Meteor (stack is mostly open-source, and closely related with the business)
Miguel@qualia.com

Dandy

Dental practice
Want to provide an integrated suite of products and tools that don’t directly provide patient value (insurance, etc)
Engineering team of about 5, team size is about 20
$30M in total

uid: 202009231700 tags: #meetings #2020recruiting

February 22, 2023

Asking Nate About Stripe

So, take all this with a hefty serving of salt. My opinions are all formed from 2 weeks of zoom/slack/email which is not necessarily the best signal.

Does Stripe feel more unified culturally?

I think it is more unified than Theorem, but Theorem felt crazy siloed in spite of the small size. AFAIK, Theorem didn’t really have a defined culture which was probably a part of this.

Whereas at Stripe, there is a defined culture (https://stripe.com/jobs/culture). Every person joining Stripe goes through an onboarding that includes all new hires for that week. This even included the new CRO which was a surprise for me.

Ultimately, sticking to culture is going to depend on individuals, but it’s hard to expect much unification without setting the direction.

At Clever (spent 4y there before Theorem), the culture tenets were defined prior to 50 employees and I think that contributed to everyone staying aligned. Different teams had different ‘weights’ to each of them, but I think we were aligned on the whole. https://blog.clever.com/2016/05/scaling-clever-culture/

My takeaway from this reflection is that you probably cannot expect a company to culturally unified unless they either:

are a small company (but this isn’t a guarantee - see Theorem)
define their culture explicitly

How does the quality/scope of work compare to Theorem?

Haven’t started any real work with my team yet.

My read from docs/messages I’ve been reading is that the scope of the work is going to be far tighter than at Theorem. However, the quality of the work (at least for me) is going to be better.

As a new grad engineer, I think the opportunities you can find will pretty consistently follow:

bigger company: smaller scope, higher weight on quality
smaller company: bigger scope, lower weight on quality

Do you feel like you have more or less ownership over the projects you’re working on?

TBD here. I’m anticipating that I’ll have limited ownership the first 6 months and will be getting opportunities to lead some projects after that point.

Created from: Asking Successful People 202009140034

uid: 202009140035 tags: #2020recruiting

February 22, 2023

Imitation Learning

In driving, observations consist of observations of car’s camera, and actions are the way that you turn the steering wheel.

You will collect a large dataset consisting of (observation, action) tuples, and you can learn supervised learning to figure out actions given certain observations. (behavioral cloning)

hi world

Behavioral cloning doesn’t work (in theory)
- Small mistakes compound into larger mistakes, because small mistake veers the training trajectory off course, which makes it more likely to make bigger mistakes, because it is not used to anything off the training trajectory.
But sometimes, it works in practice (a lot of data and a few tricks)
- Why? Can be partly explained by the fact that they used three different camera angles (left, right, forward). Left and right cameras are teaching the car how to correct mistakes, which will help it correct itself after it has veered off course
Mathematically, this “training error drift” can be explained by the fact that the training and test distributions are not the same - when you take an action, it results in different observations, which changes the distribution.

Instead of trying to be clever about the policy, we can try to be clever about the underlying data distribution (so that the distribution of observations under the data is the same as the distribution of observations under the policy). This is a technique called DAgger (Dataset Aggregation).

Dagger steps:

Train policy from human data $D = {o1, a1, …, oN, aN}
Run the policy to get a policy dataset D_pi = {o1, …, oM} (These observations come from p_pi_Theta(oT))
Ask human to label D_pi with actions a_t
Aggregate the dataset
Go back to step 1, train policy on this aggregated dataset.

This isn’t that useful in practice sometimes, because step 3 is hard.

Problem with imitation learning

humans need to provide data, which is typically finite
- deep learning works best when data is plentiful
humans are not good at providing some kinds of actions
humans can learn autonomously; can our machines do the same?

uid: 202008311422 tags: #cs285

February 22, 2023

Daily review 07-24-20

How did your goals turn out today?

Be organized.

My Things is organized at least. I generally don’t have enough time (or energy, I guess) to do enough things that are of my core list of todos anyways, so completely setting up my Things lists on a weekday at least doesn’t matter a whole lot.

Do my best on all my commitments.

Finished most of the work on the design doc for Amazon, and also worked closely with file to get a POC for the lambda functionality as well.

Be practically independent.

Nothing too much in this regard. Talked with Mulan, later at night, about long-term plans - what I’m thinking about doing after college, from a career perspective, and she said something that struck me - “The future is coming up fast.” Kind of scary. Going to

Grow grow grow.

Read a lot of blog posts - one about how morality developed through evolution and social dynamics, one about India/Jio/four corners of the internet, one about the four different types of people (passive/agressive, conformist/independent), and also a lot of posts by Linus (thesephist.com), holy shit this guy creates SO MUCH SHIT. Pretty intimidating.

Had a good social time with the Amazon group!

Be a good son/brother/boyfriend/friend

Didn’t talk a whole lot with parents, but told them that I wanted to talk about housing stuff over the weekend, so I’ll do that.

Contribute to the world

Going to work on the productivity decal stuff this weekend!

Be more forward-thinking

I’ve recognized that early in the morning on Saturdays is an excellent time for me to work on far-reaching projects and big goals that I haven’t gotten a chance to dig into yet.

Be more disciplined

I definitely know that if someone is standing at my head, or if I’m in a meeting or something, I can work for hours at a time if that’s how it works out. The thing to figure out is how to make sure I’m productive when I don’t have ny kind of supervision. Today, I played a Lott of Hearthstone, it got pretty hard to control sometimes. The other interesting thing about is that I don’t associate it with any part of my identity at all. If people asked me what I’m currently doing, my instinct isn’t to say that I’m playing a lot of Battlegrounds and spending close to 45 min per day (probably more) watching Battlegrounds as well. That’s how little it means to me personally, but I’m still spending so much time on it? Seems like there’s other things that I still have to study/take a look at.

Be physically healthy.

Tried to do an HICT, but my phone got caught in the sofa and then I got kind of frustrated, and stopped doing it.

Any other thoughts?

uid: 202007250133 tags: #journal

February 22, 2023

← Newer Page 9 of 175 Older →

Why Trump Sucks

Reflecting On A Negative Teaching Experience

Human Capital Stealth Networking Event (_)

Agora (Ryan Gibson)

Challenges

Vise (Nikki Kang)

Technical challenges

Q/A

Observe (Jacob)

Q/A

Qualia (Joel Gottsegen)

Q/A

Dandy

Asking Nate About Stripe

Imitation Learning

Problem with imitation learning

Daily review 07-24-20

How did your goals turn out today?

Be organized.

Do my best on all my commitments.

Be practically independent.

Grow grow grow.

Develop my social network.

Be a good son/brother/boyfriend/friend

Contribute to the world

Be more forward-thinking

Be more disciplined

Be physically healthy.

Any other thoughts?