Sarthak Vajpayee

Data Scientist | ML Engineer | AI Enthusiast

About Me

I am an accomplished Data Scientist and Machine Learning Engineer with 5+ years of expertise in Natural Language Processing (NLP) and deep learning. My analytical and technical expertise was refined at the University of Texas at Dallas, focusing on Data Science within Business Analytics.

With a proven track record of developing AI/ML solutions, I excel at extracting valuable insights from unstructured data, leveraging expertise in prompt engineering, GPU computing, and transformer-based architectures. I've successfully fine-tuned models like LLaMA, Mistral, BERT, and other state-of-the-art language models using Hugging Face Transformers, Bits & Bytes, and Langchain.

I am driven by a passion to apply my knowledge and skills in real-world applications, continually seeking innovative ways to leverage data for impactful results. With strong communication skills and a willingness to relocate, I'm excited to collaborate with forward-thinking organizations.

Skills

Technical Skills

Python95%
Machine Learning90%
Deep Learning85%
NLP92%
Big Data88%

Tools & Frameworks

TensorFlow88%
PyTorch85%
Scikit-Learn92%
Hugging Face90%
LangChain85%

Soft Skills

Problem SolvingTeam CollaborationCommunicationProject ManagementAdaptability

Experience

Research Assistant - The University of Texas at Dallas

January 2024 - August 2024

  • Boosted research efficiency by 10x by developing a Retrieval-Augmented Generator (RAG) leveraging Large Language Models (LLMs), LangChain, Pinecone Vector Database, text embeddings, and Python.
  • Surpassed existing state-of-the-art sentiment classification research by 20% through fine-tuning transformer models (T5, BERT, Mistral, LLaMA) using PEFT techniques (QLoRA), Hugging Face, PyTorch, and Bits & Bytes.
  • Enhanced Transformer models (BERT, DistilBERT, RoBERTa) using PEFT techniques (QLoRA, IA3) with Hugging Face in Python, achieving 40% improvement in F1 micro-score for classification tasks.

AI Software Engineer - AppSteer

May 2023 - December 2023

  • Catalyzed $200K in revenue by pioneering an Agent-based Q&A ChatBot, leveraging Large Language Models (LLMs), LangChain, Python, PySpark, and PostgreSQL to reduce error rates by 50% through expert prompt engineering.
  • Accelerated time-to-market from 4 days to 2 minutes, revolutionizing application development and deployment through automation using FastAPI, Pydantic, Docker, LangChain, Hugging Face, PyTorch, and OpenAI on Azure.
  • Enhanced deployment rate by 20% by orchestrating over 50 cloud ETL deployments, streamlining CI/CD workflows with Git version control, Jenkins, Kubernetes, and Prometheus.

Data Scientist - EY

December 2021 - July 2022

  • Drove $1.1M in annual cost savings by optimizing sales forecast models through efficient data processing with Azure Databricks and strategic feature engineering techniques with Azure ML, resulting in an 8% improvement in demand forecast accuracy.
  • Set new benchmarks in international market analytics by developing a cutting-edge, second-generation predictive system combining ARIMA and XGBoost in PySpark for advanced data modeling, achieving a remarkable Mean Absolute Percentage Error (MAPE) of 10% within 15 days.
  • Boosted forecast accuracy by 15% through meticulous time-series analysis and A/B testing, leveraging Tableau dashboards for enhanced data visualization and insights into critical demand trends.

Publications

COVID-19 Diagnostics through AI

Developed an AI-powered COVID-19 diagnostic tool that matches the accuracy of traditional tests.

AI-Driven Liver Cancer Diagnosis

Created a life-saving tool that predicts mortality risk with high accuracy, enabling clinicians to make informed decisions

Transformers (State-of-the-art NLP)

An in-depth exploration of transformer architectures and their applications in NLP.

Cracking the GRE with AI!

A case study on applying ML & NLP techniques to optimize standardized test preparation.

More Publications

Projects

Sentiment Analysis for Social Media Marketing

Created a real-time sentiment analysis tool for Twitter data, helping businesses gauge public opinion on their products

TableTalk: AI-Powered Data Analysis Chatbot

Developed a proof-of-concept chatbot using Agentic AI to demonstrate intuitive data analysis and visualization capabilities

㊟ Attention Mechanism! Japanese → English

Developed a seq-to-seq model with attention mechanism for Japanese-to-English translation, achieving a 0.57 BLEU score

Global Top 5% in Google Quest Q&A Challenge

Created an ensemble of transformer models (BERT, ALBERT, RoBERTa, XLNet) to accurately predict StackExchange QA ratings

More Projects

Education

Master of Science in Business Analytics

The University of Texas at Dallas, USA

August 2022 - August 2024

Bachelor of Engineering in Electronics Engineering

Dr. A.P.J. Abdul Kalam Technical University, India

August 2014 - June 2018

Certifications

  • More Certifications

    • Contact

    I'm always open to new opportunities and collaborations. Feel free to reach out!

    EmailLinkedInGithubResume