Lara Thompson
Yet another physicist doing ML.
Principal Machine Learning Scientist / Engineer at Tableau / Salesforce. I started out on the Explain Data feature team then I was briefly on the Einstein Discovery team. Since the summer, I've been working on Tableau Pulse developing insights that are the heart of Tableau Pulse and architecting new GAI features.
Bouldering more and more (regularly getting up to ~V3-V5; also finding harder and harder V2). Ran the Enchantments in-a-day (many hours); ran a semi-fast 10km (48:20).
My first gig outside of academia. I was hired to improve their appliance disaggregation (separating the real+reactive electrical signal of a house into the constituent appliances). My first prototype outperformed their deployed version but it was based on edge detection and edge pairing, and didn't work in many homes (e.g. with heaters that cycle on and off at high frequency and varying amplitude). A U-Net model with attention could learn to separate individual appliances well but we still needed data per appliance type, sometimes even per make and model (after heating and cooling, dryers and fridges are the biggest power consumers and both are fairly distinctive, but not always; e.g. the sharp initial spike of an old fridge cycling on is nearly completely suppressed in an ultra-efficient modern fridge).
Home electricity monitoring enables smarter solar + battery usage: I devised control algorithms for various tariff structures. E.g. Solar panels produce peak power midday while most homes have peak consumption in the evenings; storing the excess in a battery for later, when tariffs are higher, offsets the cost and better utilizes the solar generation. For all tariffs, I compared against optimal in-hindsight control to quantify regret; mostly, optimal control is possible in most homes because the battery is generally small enough it's always worth fully charging and saving for the evening.
By the end of my three years I was leading an ML team of 4 and had run a decent half marathon (1:49:29).
Back to UBC, joining Kurt Haas's neurioscience lab to do analysis work. I started attending a data science paper reading meetup that I still attend / help organize; find us on discord.
Climbed Mt Rainier (Disappointment Cleaver router) and Pico de Orizaba (with lots of fresh snow!) with MITOCers.
Postdoc with Patrick Lee looking for quantum spin glass signatures. This work involved some tricky numerics to theoretically support measured experimental signatures in a candidate quantum spin glass. MIT was such an incredible place with hallways full of science experiments, students studying and random dancing.
I taught at the winter school run by MITOC: ice climbing (that I was new to!), winter hiking and camping (in a quinzhee rather than the snow caves I'd tested in BC mountains); then lead climbing in the summer.
MSc / PhD in Condensed Matter Theory with Philip Stamp. My thesis on the "Equation of Motion of a Quantum Vortex" was deemed by D. J. Thouless (as external reviewer of my thesis) to resolve the decades-long dispute between his group and researchers in Russia: not in his favour!
Joined the VOC, taught at glacier school, fall long hike (a climbing school); took up backcountry skiing, mountaineering, and trad climbing. For a year, I played on the Junior Varsity volleyball team (as offside) before my fingers demanded I choose between volleyball and climbing. Climbed a bunch of mountains (Matier, Joffre, Rainier, Wedge), skied a few traverses (Garibaldi Neve, Spearhead in-a-day, Wapta icefield), led a few climbs (Snake, Over the Rainbow, Exasperator) and one boulder (Tumbleweed V2, though I tried many others: bouldering is hard).
Hosted on Github Pages