Samyamoy Rakshit rebuilds published ML models from the paper on one laptop — the transformer, BERT, ViT — and writes up what broke. Data scientist at Sigmoid Analytics.https://samyamoyrakshit.com/enhttps://samyamoyrakshit.com/blog/bert-from-scratch/https://samyamoyrakshit.com/blog/bert-from-scratch/A 7.5M-parameter Bengali BERT, pre-trained from nothing on a laptop, that scores above published mBERT and IndicBERT on Bengali news classification.Sun, 09 Aug 2026 00:00:00 GMTBERTPyTorchBengali NLPhttps://samyamoyrakshit.com/blog/attention-from-scratch/https://samyamoyrakshit.com/blog/attention-from-scratch/The original transformer rebuilt component by component, then trained to translate English into Bengali on one laptop.Sat, 08 Aug 2026 00:00:00 GMTTransformerPyTorchMachine translationhttps://samyamoyrakshit.com/blog/vit-from-scratch/https://samyamoyrakshit.com/blog/vit-from-scratch/A ViT trained on 1,935 photographs of Bengali terracotta temples — a dataset I had to build first — to test the paper's claim about pre-training. It failed, exactly as predicted.Fri, 07 Aug 2026 00:00:00 GMTVision TransformerPyTorchComputer vision