01 / 05 — START WITH A REAL PROBLEM
A clear reason to notice someone.
Shopify’s July 2025 catalogue article reported 40 million daily requests to AI models to organize product information. Running a model is called inference; at that scale, response time and computing cost matter.
Tri Dao, Together AI’s co-founder and chief scientist, coauthored FlashAttention, which makes a key AI calculation more efficient on GPUs, the processors used for this work. Constellation shows why the company challenge makes this research worth exploring, then links it to its contributor. Following that chain gives a recruiter specific evidence to discuss, beyond a reputation or job title.
The evidence is public; the proposed technical fit remains a hypothesis.