Tom Almog

cs @ waterloo. building ML systems.

now

• founding engineer at Parkalytics. building models and systems for drone data collection.
• building Peira. RL environments for physics and engineering.

before

• Polymarket (apr 2026) - autoencoder flagging anomalous trades for insider trading review.
• WatStreet (dec 2025 to mar 2026) - transformer time series models for market making.
• Wat.AI (sep to dec 2025) - conducted research on retinal lesion segmentation, built the best model for diabetic retinopathy detection
• Waterloo Aerial Robotics Group (may to aug 2025) - perception models running on real drones at 45 FPS.

things i made

• Crucible - agentic platform for fine-tuning and evaluating LLMs.
• Tempo - rust CLI that runs Claude Code overnight. 9,000+ downloads.
• CarbonAware-ML - pauses PyTorch training on live carbon intensity signals. 2,400+ downloads.
• an analysis of optimizer choice on energy efficiency - first author paper, under review at Sustainable Computing.
• WagerLoo - prediction market for waterloo co-op salaries. 80+ student sign ups.

hackathons

• 1st @ NexHacks, polymarket track
• 1st @ LAHacks, fetch.ai track
• 1st @ HackBrown, fetch.ai track

want to reach out? email me here talmog@uwaterloo.ca

• github.com/tomalmog
• linkedin.com/in/tomalmog