Less by handmore by model
As an AI/ML intern at Fixo (IIT Madras), I built and deployed ML models for predictive analytics and automation. The work they took over cut the team's manual processing time by 75%.
Read this firstThis one is about a team getting hours back, not a leaderboard score. The model only counts if it's in the workflow.
Business data to an automated workflow
The models sit inside the automation workflow, so their output replaces a manual step instead of a report someone has to read. Click a stage to read what happens there.
From a person's queue to a model's
Before, every item went through someone's hands. After, the models handle the routine prediction work and pass the result on automatically. Switch between the two.
Dashed boxesThe parts I haven't written up yet. The shape is right; the labels come next.
DraftName the data source, what the models predicted and how they were served.
Three quarters of the hand work, gone
Each square is 1% of the manual time the process used to take. After deployment, 75 of them are handled by the models and 25 still need a person.
15 h back every week
Honest readingThe workload is yours to pick. Only the 75% cut comes from the project; the hours are plain arithmetic on top of it.
DraftSay how the 75% was measured: which tasks, timed before and after, over how long.
Six stages, end to end
The same six stages as the homepage. Here the last one is the step that matters most: the models running inside the team's day-to-day work.
Why it mattersPlenty of models never leave a notebook. This one changed how much of the day people spent on manual work.