Career journal for

ML Engineers

ML work is half experimentation that did not produce and half production work that compounds quietly. By review time the journey of the 14 experiments that did not work is hard to defend.

Bloom is the career journal built for the moments the work actually happens, not the night before the review.

A captured day for a ML Engineer

1. 01

   9am: shipped the v3 ranking model to production. Online numbers tracked clean for 72 hours.

2. 02

   12pm: data review with the platform team on the new feature store; my concern about backfill latency changed the design.

3. 03

   3pm: paired with Marcus on his first feature-engineering project.

4. 04

   5pm: wrote the eval-framework doc that 3 other ML teams have started using.

What a ML Engineer captures

• Model shipped to production with measurable lift
• Inference latency and cost wins
• Feature engineering and data-pipeline work
• Evals and offline-eval framework contributions
• ML-platform tooling

Shipped ranking-model v3 (5.4% online CTR lift). Cut inference cost $180K annualized. Authored eval framework adopted by 3 ML teams.

Promotion rubric, mapped to capture

• Model performance outcomesCaptured automatically through dated entries, auto-tagged against this dimension, and surfaced in your generated Performance Report and Period Recap.
• Production reliabilityCaptured automatically through dated entries, auto-tagged against this dimension, and surfaced in your generated Performance Report and Period Recap.
• Data and feature engineeringCaptured automatically through dated entries, auto-tagged against this dimension, and surfaced in your generated Performance Report and Period Recap.
• Cross-team partnershipCaptured automatically through dated entries, auto-tagged against this dimension, and surfaced in your generated Performance Report and Period Recap.