hi, i'm sarosh doddaka.
i build software and machine-learning tools for real-world problems, and hope to one day back ventures that bring useful solutions to more people.
build what works. scale what helps.
right now, i lead task intelligence at conductflow and develop computer-vision models at northwestern kellogg.
some things i've worked on:
- conductflow: i build the model-inference layer. across 10+ pilots, 73% of outputs were accepted without edits; teams reported 40-70% less task time.
- northwestern kellogg: date-label detection at 97% accuracy across thousands of images, in 0.33 seconds per image.
- scribbleitdown: a digital notebook for first robotics, used by 1,000+ ftc participants.
- innovateatl: co-founded the organization, building atl's hs startup pipeline (200+); supported by tech square ventures, create-x, and microsoft.
- prev. deal team @ 3lines, on the deal screening team. screened 100+ startups, modeled financials and pricing, and presented opportunities to gps.
also: auralis (portable respiratory disease detection); podcast co-founder (1m+ views).
recognition: 3x fbla national finalist; ftc inspire and innovate award winner; first dean’s list semifinalist; georgia first symposium presenter (1 of 4 from 2,000+).