Person peeking from behind laptop
Available for Full-time Work

Scalable Neural Models for Your Needs

I craft domain-guided neural models using advanced machine learning and natural language technologies.My work blends accuracy and interpretability to provide data-driven insights for impactful outcomes.

About Me

Soumyajit Gupta

Machine Learning - Neural Nets - Natural Language - Computational Sciences

This is Soumo, based in Reno, Nevada.
I'm a representation and interpretation-driven neural modeler and data scientist.

I recently finished my PhD on Multi-Task models for group-targeted Toxicity detection, co-advised by Matthew Lease and Maria De-Arteaga, from UT Austin, Department of Computer Science.

I am currently available for Full-time Remote or In-person positions in the Reno and Las Vegas area.

profile image
Performant
High-Precision
Scalable
Low-Weight
Interpretable
Domain-Guided
Reliable
Deployable
Usable
Performant
High-Precision
Scalable
Low-Weight
Interpretable
Domain-Guided
Reliable
Deployable
Usable

Real-World Results

Featured Projects

Explore my journey of shaping ideas into practical and scalable outcomes

UT Austin•2021

Neural SVD solver for Big Data

  • Two stage neural engine as alternative to randomized SVD techniques
  • Explicit Memory requirement: guided by feature dimension and desired rank
  • Fully interpretable model: all outputs and weights have specific meaning
Neural SVD solver for Big Data
UT Austin•2023

Multi Task Learning Toxicity Model

  • Conditional MTL model to learn toxicity targeted at different groups
  • Improved Recall ~8% and ~15% over Independent and SoA MTL models
  • Runtime and Parameter reductions by ~56% and ~72% over Baseline
Multi Task Learning Toxicity Model
UT Austin•2025

GAP for Target-group detection

  • Group-fairness loss function based on Accuracy Parity measure
  • Balanced group accuracy around Target-group detection
  • Group disparity reduced from ~22% to ~8% with minimal accuracy drop
GAP for Target-group detection
Performant
High-Precision
Scalable
Low-Weight
Interpretable
Domain-Guided
Reliable
Deployable
Usable
Performant
High-Precision
Scalable
Low-Weight
Interpretable
Domain-Guided
Reliable
Deployable
Usable

My Portfolio

Education

PhD - UT Austin

Computer Science

2018-2024

MS - UT Austin

Computer Science

2014-2017

MTech - IIT KGP

Electronics Engg.

2012-2014

My Tools

Tech Skill Suite

Technologies that power my projects

Programming Languages

Python
C
C++
CUDA
Visual Basic

Frameworks

Tensor Flow
Keras
PyTorch
Dask
PySpark

Libraries

Numpy
Pandas
Scikit Learn
Scipy
XGBoost
OpenCV
Hugging Face
Spacy

DBMS

MongoDB
Postgres

Front-End

ReactJS
NextJS
Tailwind CSS
HTML

LLMs

Bert
ChatGPTGPT
Llama
Mistral

Others

Github
Rest
Docker