Vikram Oberoi
2.7K posts
@voberoi
Engineer @ https://t.co/7A7etCio2f. Creator @ https://t.co/cQzPDZPwdf. Writing @ https://t.co/0tnGGKTyGi. Curator @ https://t.co/DXRUIzbyHk.
We shipped Chandra OCR 1.5 to the Datalab API this week! This is a big improvement over 1.1, especially on tables, lists, charts, chemistry, and math.
I'm excited to open source Chandra OCR 2! - 85.9% (sota) on olmocr bench - 90+ language support w/benchmarks - 4B model (down from 9B) - Full layout information - Extracts + captions images and diagrams - Strong handwriting, math, form, table support
I'm excited to open source Chandra OCR 2! - 85.9% (sota) on olmocr bench - 90+ language support w/benchmarks - 4B model (down from 9B) - Full layout information - Extracts + captions images and diagrams - Strong handwriting, math, form, table support
We're on drought watch here in NYC. Here's a bit about where our water comes from and something we have called a "Filtration Avoidance Determination" (FAD) we have with the EPA. (This is the New Croton Reservoir. It's pretty and you can go there.) This is a 🧵 --
Another departure today from NYC agencies — @NYCWater — DEP Commissioner Rit Aggarwala resigned. No word yet on a replacement from @NYCMayor admin. #NBC4NY
claude code isn’t factorio — it’s dwarf fortress
Happy New Year! A quick update: @citymeetingsnyc will be back in early 2026. It’s taken me longer to figure out how to give it the time it needs w/ my day job @datalabto but I’m looking forward to getting it online again. More to come.
New Year, New Case Study 💼 Cofactr (YC W’22) is a source-to-pay and 3PL platform for critical hardware manufacturers, supporting aerospace, defense, robotics, and medical devices. They’ve raised $28.8M from Bain Capital, Y Combinator, and others, and support customers like Amazon’s robotics division, operating in environments where document accuracy directly impacts production.
