brian_spiering

Location: San Francisco, CA

Remote: Yes

Willing to relocate: No (Open to travel)

Technologies: NLP, ML, DL, Python

Résumé/CV: https://docs.google.com/document/d/e/2PACX-1vTcDv3aSWg6CLNBn...

LinkedIn: http://www.linkedin.com/in/brianspiering/

Email: bspiering@gmail.com

Bio: I'm a full stack data scientist, specializing in NLP. 9+ years of experience with a wide variety of projects and organizations. I have led technical teams to go from business idea to deployed model.

Right now, I'm freelancing. Open to part-time and full-time roles.

I'm also a technical trainer for machine learning and Python.

SEEKING WORK | Remote | SF, CA | ML Engineer

Hi Y'all, I'm Brian. I'm a full-stack ML Engineer - from business use case to deployed machine learning model.

Working in Artificial Intelligence (AI), Deep Learning (DL), and Natural Language Processing (NLP).

I also provide training services to level-up engineers in AI and ML.

Please get in touch bspiering@gmail.com or http://www.linkedin.com/in/brianspiering/ Would love to connect to chat in more detail!

I choose not to focus on the differences between Bayesian statistics and machine learning. I choose to integrate the best parts of both.

If you are curious, the book "Probabilistic Machine Learning: An Introduction" by Kevin Murphy goes into greater detail.

I did this for a course I taught that involved mostly coding. The student watched video lectures before class and took a quiz to make sure they understood the content. In class, I answered questions, showed how to do sample activities by live coding, and the student did activities either individually or in pairs. I gave real-time code reviews while they were working.

I got pushback from the school for using someone else's videos (they really wanted me to record my own video lectures AND teach a live class). I also got pushback for not failing more students. The previous teacher just lectured, never live coded, and never did any code reviews for students. The "active learning" instruction was more effective than lecture based instruction.

Agreed! My 95lb German Shepherd / White Swiss Shepherd chases away packs of coyotes and bears.

I made a similar transition. I suggest finding any role in a fast growing organization and then working your way up.

There are EdTech specific job boards like http://www.edsurge.com/jobs

I used breakout rooms for small group activities. The activities were mostly discussion and solving problems. I would hop from room to room to help out (and make sure student were on task). After, I would bring all these student back together and randomly call on a student to present.

I used a flipped classroom with video lectures created by someone else so these breakout room activities could be over 50% of the time.

Newton the Alchemist: Science, Enigma, and the Quest for Nature's "Secret Fire" by William Newman details Isaac Newton's long-term studies in alchemy.

The CS-Ed Podcast has an episode that goes into detail on how UC Berkeley’s EECS undergraduate teaching pipeline works. https://csedpodcast.org/blog/season3_episode3/

Thank you for the tip. The partnership is called "O'Reilly for Public Libraries". I just logged onto the full O’Reilly Learning platform through my San Francisco public library account.

One area to explore is how to optimize a sequence of events (e.g., user journey). Reinforcement learning (RL) might be useful for that.

The increased role of computation and machine learning in many scientific fields is a paradigm shift.

You are correct. A better word is amortality - a lifestyle that defies the process of aging.

I found running a Getting Things Done® (GTD) style system is useful - Write down what is on top of mind, make a complete list of all projects, do items that are contextually relevant, and at least once a week review everything.

The physical behaviors are more important any specific technology.

The critical piece for us is a new fresh whiteboard for each day. Write down all the items in play so everything is visible at once. Then prioritize. Most importantly - go do!

As a former professor, I want to second this advice. Even asking the professor to re-explain some from class is often helpful. If you run out of things to ask, ask about their research. Building that relationship will help if you need an extension on assignment, a bump in a grade, or a letter of recommendation later.

Some professors won't care and it will be of little value. Do not get discouraged if it is not a positive experience. Find the professors that do care.

AI Engineer

What works for me is setting a schedule and having milestones. I follow a plan for a period of time or until I create a concrete deliverable. Then I decide to work more on the project or pause to peruse something else.

I found constantly deciding what to do was too exhausting.

The article is high-level musing.

It would be useful to also see specific examples from models trained on data.

I think about technical skill development as a diverse portfolio. I spend a majority of learning budget on fundamentals, mostly overlearning core knowledge and skills. I spend a small amount of my learning budget on specific techniques, tools, and packages that currently popular. A very small fraction of my learning budget is spent on high risk / high reward elements.

Large consulting firms (e.g., McKinsey & Company, Bain & Company, and Booz Allen Hamilton) often look for broad experience and aptitude.

UCSB is very close to the Pacific Ocean and much of the ground is sandy cliffs that frequently fall into the Pacific Ocean. The university is trying to build for the long term and is very concerned about the possible impacts of climate change that might accelerate erosion. Those factors contribute to increase building costs.

The highest value projects to employers are individual, custom, novel, contemporary, domain-specific, applied, potentially high-value business impact, and end-to-end.

Lower value projects are the inverse - group, generic, templated, uses outdated technologies, rehash of common examples, academic, and research-oriented.

One option might be The Algorithm Design Manual by Steven Skiena and the associated website https://www.algorist.com/.

It is more problem-solution based with applied examples.

The issues you describe are important but typically out of scope for data structures and algorithms book. They are often considered more architectural issues.

In many real-world scenarios, it is not necessary to pick the best data structures and algorithms. It is often good enough to just avoid poor choices.

Trustworthy measurements are important since constants and implementations matter. That is done through benchmarking and profiling. This is also often not considered part of many data structures and algorithms books.

Bayesian hypothesis testing, including Bayes factors, might be more useful.

I have worked at many bootcamp-style educational programs.

Teaching in a bootcamp-style environment is a team endeavor. Recent graduates can be a part of that mix.

Generally, it makes sense to hire some of the instructor pool out of recent graduates. It is often difficult to find instructors at all. The recent graduates know the curriculum. It also helps those recent graduates develop a deeper understanding of the material and extend their job searching window. It can be a form of peer learning that some students respond well to.

There are also definite limitations. Primarily, the recent graduates might not have great teaching skills, knowledge outside of the curriculum, and knowledge of real-world applications.

Related to starting in tech support, try to get an entry-level job in quality assurance (QA) / quality control (QC). You'll get exposure to software engineering and software engineers, then make more an informed decision about the field.

I teach others.

It forces me to understand the material because I have to actively recall it. I invent multiple ways to explain the concept which deepens the encoding.

I design and solve exercise of increasing difficulty.

Basecamp has “summer hours” - May 1 through August 31 has 32-hour work weeks (Monday-Thursday 8-hour per day) https://m.signalvnoise.com/employee-benefits-at-basecamp/

That might be a big part of the business context behind the "We are going to be Earth’s Best Employer…" statements in the 2020 Letter to Shareholders.