occasional engineer, wannabe pro-skier @canadaski
Joined March 2024
- Tweets 229
- Following 885
- Followers 357
- Likes 15,048
59 Photos and videos
Pinned Tweet
Apr 7
We won the @HockeyAnalytics Big Data Cup 2026!
Forechecking is usually described as a system label. We tried to show it as a spatial timing problem: who arrives first, which lane is open, and how fast F2/F3 take that lane away.
15
32
393
40,836
Luke retweeted
May 20
Victor Wembanyama dribbled a basketball up a dangerous mountain route in 4.5 hours (takes a normal person 7-8 hours to walk up it) while training with the monks, per @ramonashelburne
“One day he told Wembanyama to dribble a basketball up another dangerous mountain route to Sanhuangzhai, a monastery deep in the Song Mountains. The hike traversed cliffside plank paths, suspension bridges and ancient forests, and was five times as long as the one to Bodhidharma Cave. The trail forces you to climb roughly 2,500 feet in elevation across uneven ridges and stone.”
(espn.com/nba/story/_/id/4881…)
270
1,738
27,432
6,634,366
Luke retweeted
Been drinking pretty much only Diet Coke to save water for Claude
235
4,747
81,554
2,004,909
Apr 7
Shoutout xgboost!
Mar 19
Last night I ran a March Madness ML workshop with my school’s AI and Sports Analytics clubs. Wanted to share the top 10 teams from the models we built out, along with their probabilities of reaching each round.
Repo: github.com/lblommesteyn/marc…
1
15
3,207
Apr 7
We won the @HockeyAnalytics Big Data Cup 2026!
Forechecking is usually described as a system label. We tried to show it as a spatial timing problem: who arrives first, which lane is open, and how fast F2/F3 take that lane away.
15
32
393
40,836
Apr 7
One of the main takeaways was that strong forechecks are layered forechecks. F1 matters, but the best denial sequences were the ones where F2 and F3 arrived in time to remove the next options and compress the play.
2
2
31
1,920
Apr 7
Our framework also outperformed simpler proximity-based baselines in turnover prediction. Really proud of this work and Ryan, Eric, and Yashna for working on this with me! #BigDataCup #HockeyAnalytics
Report Code: github.com/lblommesteyn/bdc_…
1
4
25
2,403
Mar 25
Are there any settlers of catan goats in my network I can run some ideas by 🤔
2
394
Luke retweeted
Mar 23
see how 4 university students reverse engineered Google's most advanced AI chip
tomorrow
13
41
646
45,829
Mar 19
Wanted to make my portfolio a little more fun, so I added Pokemon-inspired pulls. Pull to check out my experience, or pull real tcg cards at lukeblom.fyi
6
15
747
Luke retweeted
Mar 19
me watching a 16-seed up 10 points on a 1-seed: "oh no this is going to kill my model's brier score"
2
1
2
553
Mar 19
Last night I ran a March Madness ML workshop with my school’s AI and Sports Analytics clubs. Wanted to share the top 10 teams from the models we built out, along with their probabilities of reaching each round.
Repo: github.com/lblommesteyn/marc…
1
1
22
4,051
Mar 19
Wanted to share that my team finished top 3 at the SPLxUTSPAN Data Challenge🏀
we predicted free throw landing spots using only maskless 3D pose data with a physics-Informed bayesian calibration approach
down to 0.0068 MSE. ~2 inches of error. 1/2 🧵
Jan 23
Been working on some inferred projectile motion gifs for the @mlsepr SPL x UTSPAN 2026 Data Challenge
I've still got a few spots left on my team so if you're a canadian student that knows a lot about ml and or biomechanics and wants free raptors tickets pls dm 🙏
2
1
12
1,160
Mar 19
shoutout to the MLSE SPL, UTSPAN teams for putting on such a fun comp 🙏and @_ousac for letting us present in front of a huge crown!
and to my teammates Yuvraj & Nathan, couldn't have done it without you guys
full writeup below if you're interested 2/2
github.com/lblommesteyn/spl_…
2
179