ASHUTOSH MISHRA / JOURNAL
Ideas in progress.
Lessons from building.
Notes on machine learning, language, and the tools behind the work. A record of questions, experiments, and things learned along the way.
From the archive
11 entriesMachine Learning / Projects
Zephyr: Exploring Shape and Airflow with AI
Building an aerodynamic design workspace that connects plain-language requests, bounded geometry changes, and NVIDIA DoMINO predictions in one inspectable comparison.
Machine Learning / Benchmarks
What I Learned Benchmarking Jev Against Classical Machine Learning
I tested Jev 1.13.0 against eleven classical classification pipelines on eight datasets. The interesting result was not a universal winner, but a clear map of where each approach worked.
Hackathons / Machine Learning
Rune Goblin: My Build Small Hackathon Winner
My winning Hugging Face Build Small project: an AI dungeon crawler that turns hand-drawn runes into spells, with a fine-tuned vision model and a deterministic game engine.
Deep Learning / Python / Machine Learning
GPU Server SetUp for DeepLearning
Gcloud server setup
tech
Pytorch English to Hindi Translation using Attention
Using encoder-decoder model with Attention network to create a Langaue translation model
Deep Learning / Python / Machine Learning
Uncertainty Quantification
UQ
tech
Pytorch Mnist CNN
CNN on Fashion Mnist using Pytorch
Deep Learning
Doodle GAN
Doodle GAN project
Deep Learning
Deep Representation of Visual Decriptions
Final Year Project
Deep Learning
Emotion Detection
Emotion Detection using bi-CNN
Deep Learning / Geo-Loc
The BOT
Smart BOT
CONTINUE EXPLORING