Six weeks is a very short period for a ~25 person team to come together, reconcile goals and expectations, and produce results. Consequently, we are constantly on our toes, changing directions of research and taking risks as needed. Every day is different, and our week-to-week schedule is irregular, too. Despite this, I’ve fallen into somewhat of a routine here at JSALT.
Daily Schedule
In general, my days look a little something like this…
8am: I wake up, have some breakfast and listen to music at my apartment.
I check whatever process I have running in the morning because it’s likely to have failed the night before.
Around 8:45 I take the ‘scenic route’ from my apartment to the ETS building for work. In reality, it takes about four minutes to walk from door to door. I take about 15 minutes, sometimes making circles around buildings just to spend some more time in the warm sunshine and fresh air before I’m holed up in the lab all day.
9am: The team has a standup meeting. We stand in a circle and give a short blurb about what we were doing the previous day and if we have any ‘red flags’ that prevent us from moving forward with our research.
We work in one of ETS’s classrooms. We each have our own workstation, and some have monitors. To practice presentations (which we do often), we just slide down the projector and present to the whole room.
In the beginning, we agreed to have meetings outside of the classroom so as not to disturb others. As time went on, though, we became quite lax with the rule so usually there’s some chatter. I prefer it that way – I don’t like everyone being together but not interacting, staring at the computer screens. Even for small questions, I’ll bypass a Slack message to go and talk to a teammate. Every couple of hours my mentor, Paola, comes to check in and make sure everything is going smoothly.
12pm: I stop by the grocery store, which is in the building adjacent to my apartment and the ETS work building. I’ll get some bread and maybe some fruit and veggies, and then eat lunch in my room. If I feel like it’s too quiet, I’ll put on a Stuff You Should Know or Throughline or Planet Money podcast. They keep me good company.
1pm: Walk the short distance back to our workspace, again taking the scenic route if I’m wanting some sunshine.
3pm: We have an inter-team ‘cookie break’ that is meant to facilitate interaction between the five teams. I was skeptical of the function at first, but I think it helped to improve relations between the teams even if we didn’t adhere to strictly work-related conversations. I did hear some idea exchange and brainstorming, though.
6pm: In the early weeks I was going to play beach volleyball most evenings, just 5 minute walk from the apartment. Recently, I have been running along Canal Lachine.
8pm: Cook dinner. I’ve tried some interesting recipes but for the most part I stick to basic carb + vegetable meals. My most recent discovery is blue cheese and pear on toasted baguette.
9pm-12pm: Back to work (at my desk in the apartment) to debug or study statistics or prepare slides.
Memorable milestones week-to-week.
Week 2. I presented my work up until that point at our weekly undergraduate lunch meeting. I spoke for about 15 minutes and then answered a few questions. My presentation was an overview and introduction to the diarization problem:
Week 4. On Wednesday, Yoshua Bengio gave a plenary talk on abstracting data into useful high-level representations. He may be the only Turing Award recipient I’ll ever get to hear speak in such an intimate setting.
Thursday, I had another undergrad lunch presentation. Before the lunch, Xuedong Huang – who leads Microsoft’s Speech and Language Group – gave a plenary talk. He then expressed interest in attending the undergrad presentations, so I got to present to him. It was certainly a privilege to do so, but I was nervous. It was a little unreal to be giving a presentation to someone in his position.
We walked to the Google offices downtown Thursday evening for a get-together with a few (turns out, very few) of the employees.
Week 5. I got to have lunch with my (current) top-choice graduate school professor, Dr. Nima Mesgarani. He heads the Neural Acoustic Processing Lab at Columbia U.
Week 6. Final presentations and lunch with Microsoft Principle Applied Scientist Jinyu Li, who gave us insight into Microsoft’s data policies and the speech/language team’s work philosophy.