Open to research & data roles · Queens, NY

PareshSubedi

I build machine learning systems that think efficiently.

M.S. Data Science, Fordham University. Research in efficient Transformer models, and applied ML, data engineering and analytics.

See the work Resume

Scroll

01 — About

Iworkwhereresearchmeetsproduction:makinglargemodelsefficient,andmakingdatatrustworthyenoughtoacton.

Graduate Research Assistant at Fordham University, researching neuroscience-inspired ways to make Transformer-based LLMs faster and more energy-efficient, through inhibitory mechanisms and sparse activation. Before that I spent two years at Genpact in Bangalore, managing SAP ERP accounts and building the Tableau reporting behind them.

Validation accuracy

CNN ensemble on HAM10000

Rows engineered

A decade of daily S&P 500 data

Top of cohort

Merit scholarship, Jain University

02 — Experience

Where I've been building

2025 — Present

New York, USA

Graduate Research Assistant

Fordham University

Researching neuroscience-inspired methods to improve the energy efficiency and inference speed of Transformer-based LLMs, under faculty supervision.
Exploring inhibitory mechanisms and sparse activation to cut redundant computation in large-scale neural networks.

2023 — 2025

Bangalore, India

Senior Associate

Genpact

Onboarded and maintained customer accounts in SAP ERP systems, keeping data entry accurate and timely.
Resolved account discrepancies with cross-functional teams, reducing data errors and improving customer experience.
Built Tableau reports on customer account metrics that surfaced patterns improving operational efficiency.

03 — Selected work

Things I've shipped

01Auto-assignment

AI Agent for Ticket Triage & Auto-Assignment

A CrewAI multi-agent system that scores tickets by severity, priority and effort, then assigns each one to the best-fit developer by expertise, workload and history.

CrewAI
Multi-agent
Streamlit

0285% val. accuracy

Skin Disease Detection: CNN Ensemble

Transfer learning across DenseNet201, VGG16, InceptionV3 and InceptionV2 on HAM10000, combined through ensemble methods.

Transfer learning
Ensemble
CNN

031.3M rows

S&P 500 Market Analysis & Tableau Dashboard

A Python ETL pipeline (yfinance, BeautifulSoup) that builds a 1.3M+ row dataset, visualised in an interactive Tableau dashboard of volatility and drawdowns.

Python ETL
yfinance
Tableau

04Causal inference

Causal Factor Discovery in S&P 500 Returns

Identifying the causal drivers of S&P 500 returns with Double Machine Learning, LSTM and Random Forest.

Double ML
LSTM
Random Forest

04 — Toolkit

The stack, in orbit

Languages

Python · SQL · R · NoSQL · Bash

Machine learning

Scikit-learn · XGBoost · Pandas · NumPy · Matplotlib · Seaborn

Deep learning & AI

PyTorch · TensorFlow · Keras · Hugging Face

Big data

PySpark · HDFS · MapReduce

Tools

Power BI · Tableau · Git · Salesforce · SAP · Pega

PythonSQLRNoSQLBashScikit-learnXGBoostPandasNumPyMatplotlibSeabornPyTorchTensorFlowKerasHugging FacePySparkHDFSMapReducePower BITableauGitSalesforceSAPPega

05 — Education

Study & training

2025 — Present

M.S. in Data Science

Fordham University

Graduate Research Assistant and Graduate Assistantship recipient.

2019 — 2023

B.Tech (Honours), Computer Science – Data Science

Jain University

Coursework in Python, data analytics, machine learning, deep learning and NLP.

Recognition

Awards & honours

Graduate Assistantship
Fordham University
Competitive assistantship supporting tuition and research.
2025 — Present
Best Final-Year Project
Jain University
Undergraduate research in deep learning for dermatology.
2023
Merit-Based Scholarship
Jain University
Awarded for academic performance, Top 10% of cohort.
2019 — 2023

06 — Contact

Let'sbuildsomethingthatthinks.

pareshsubediusa@gmail.com +1 (929) 224-8329 Resume (PDF)

LinkedIn ↗GitHub ↗Queens, NY