I'm Akash Chauhan — I build machine-learning systems and backend automation tools that find structure in messy, real-world data. I care less about the buzzwords and more about the numbers a production system actually moves.
My work starts where most projects skip ahead: understanding the data. Before training anything, I run the exploratory analysis, test which features are statistically doing the work, and project the feature space with PCA and t-SNE so I know what the model is actually learning.
That habit comes from research. At IIT Guwahati's SPIN Lab, I rebuilt a computer-vision classification pipeline from feature engineering up — and pushed multi-class accuracy from 52.5% to 64.7%. The same discipline shows up across my projects: a fake-news classifier at 94%+ accuracy, deployed and live.
I'm currently a third-year Data Science & AI undergraduate at IIT Guwahati and a Python Intern at Delphic Global, focused on shipping code, automation, and models that hold up against real numbers.
Applying engineering discipline and machine learning to build scalable, data-driven systems.
End-to-end builds across NLP, tabular regression, and applied physics — each judged by a number, not a vibe.
An end-to-end detector trained on 6,000+ real-world articles. Text cleaning, stopword removal, and TF-IDF feature extraction, comparing Logistic Regression, Naïve Bayes, and Random Forest — then shipped as a live Streamlit web app.
A sales-forecasting model on real transactional data. Full cycle of cleaning, feature engineering, and EDA — comparing Linear Regression and Random Forest, with Random Forest winning on non-linear patterns. Surfaced seasonal trends and country-wise revenue distribution.
Predicting the stiffness of lattice structures from geometry — density, thickness, cell size. Built a synthetic dataset to model real material behaviour, trained Linear Regression and Random Forest, and used feature importance to identify the drivers of stiffness. An ML lens on mechanical metamaterials and biomedical design.
Open to data-science and machine-learning internships and roles. The fastest way to reach me is email — I reply quickly.