Skip to Main Content

Penumatcha Jaya Surya (Solo)

Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Penumatcha Jaya Surya (Solo) — Demo video

Ever chipped the edge of a table or desk? The usual fix is wood filler that never quite matches. A 3D printer could make a perfect matching piece, but only if you can design that broken shape in CAD software, which most people can’t. PartSnap does the design for you. Take a few photos of the chip with your phone, measure three distances with a ruler, and PartSnap gives you a 3D file you can print. AI handles the hard parts: it checks your photos, outlines the damage, designs the piece, and makes sure it’s the right size before you print it. It’s a fixed pipeline with an agent decision at each stage. Muse Spark Open (on Baseten) checks each photo and later independently reviews the finished part. Muse Spark 1.3 diagnoses the damage, can ask for one extra photo, and picks the measurements. SAM 3.1 outlines the chip in both views. Muse Wonder writes CadQuery code for the part. Plain code then verifies it: watertight, size within 0.3 mm, no overlap with the table, gap within 0.05 mm on every face, minimum thickness 1 mm. Failures go back to Wonder for up to 3 repairs. On our test chip, Wonder’s part passed every check on the first attempt (gaps 0.198–0.200 mm at 0.2 mm clearance). The first real print matched the chip’s shape but did not seat perfectly; hand measurements and photo angle are the likely causes, which a fit-test slice would address next.

Team