Machine Learning Data Scientist

Ajay Sreekumar

Reno, NV +1-775-***-**** *************.*****@*****.*** LinkedIn Website

Summary

Data Scientist with 6+ years of experience in machine learning, deep learning, and statistical modeling across research and industry environments. Proven track record of developing AI-powered solutions that achieve 90% accuracy improvements and published research in top-tier conferences.

Skills

Programming: Python (Pandas, NumPy, Scikit-learn), SQL, R Machine Learning: Deep Learning, Statistical Modeling, NLP, Computer Vision, A/B Testing Frameworks & Tools: TensorFlow, PyTorch, Apache Spark, Jupyter, Git Visualization & Analytics: Power BI, Tableau, Statistical Analysis Infrastructure: AWS, ETL Pipelines, SQL/NoSQL Databases Experience

Data Scientist Intern, NASA Jet Propulsion Laboratory, Pasadena, CA Jan 2025 – May 2025

• Joined a team of 6 to build an AI-powered copilot that automated OML code generation for NASA JPL’s systems engineering workflows. Spent 40% fine-tuning LLMs with LangChain and Python, 30% optimizing inference pipelines with advanced preprocessing, and 30% building a RAG system using Sentence Transformers and cosine similarity.

• Integrated the solution into a VS Code plugin. The result was a 90% improvement in code accuracy, 80% reduction in latency, and 30+ hours/month saved in manual engineering effort. Data Science Research Intern, Los Alamos National Laboratory, Los Alamos, NM May 2024 – Aug 2024

• Joined a team of researchers to evaluate 12 open-source LLMs (7B-34B parameters) for legacy Fortran-to-C++ code translation. Spent 50% of time designing a CodeBLEU-based evaluation framework using Python and NLP techniques, and 30% building a compiler-integrated feedback loop to improve prompt engineering for high-performance computing environments. The result was a measurable increase in translation accuracy and direct integration into HPC code migration workflows.

• Dedicated 20% of time to fine-tuning Vision Transformers for temporal forecasting on multivariate scientific data using TensorFlow and computer vision methods. The result was a contribution to a multimodal foundation model in develop- ment and co-authorship on a NAACL 2025 Best Paper Award-winning publication, presented to 100+ researchers and stakeholders across the national lab network.

Application Development Analyst, Accenture, Mumbai, Maharashtra, India Jun 2018 – Jun 2021

• Managed 12 financial applications for Mondelez International used by 50+ daily users, spending 60% of time resolving critical incidents and change requests using SQL Server, .NET, and SAP. Spent 25% applying NLP and data science techniques to analyze performance logs and client feedback, and 15% building Power BI dashboards for reporting.

• Built real-time analytics solutions across SQL/NoSQL systems by designing ETL pipelines with Apache Spark and Kafka. Resulted in scalable multi-app data workflows that reduced latency in reporting and improved decision-making across finance teams.

Projects

Medical Vision-Language Models Research 2025

• Developed custom Vision-Language Models for surgical applications using PyTorch, implementing TP-SIS models for instrument segmentation and building multimodal analysis pipelines for clinical workflows. NYC Taxi Anomaly Detection System 2024

• Implemented hybrid machine learning approach for anomaly detection in NYC taxi data, achieving 93% precision using ensemble methods and unsupervised learning techniques. Neural Network Trajectory Prediction 2024

• Built custom neural network implementation for trajectory prediction with optimized gradient descent algorithms and advanced regularization techniques using deep learning frameworks. Education

M.S. Data Science, University of Arizona, Tucson, AZ Aug 2023 – May 2025 GPA: 4.0/4.0

B.E. Electronics & Communications Engineering, NMIMS University, Mumbai, India Aug 2014 – May 2018 Certifications & Awards

• Best Paper Award, NAACL 2025

• Distinguished Scholar Award, University of Arizona

• TensorFlow Developer Certificate

• Google Data Analytics Specialization

Publications

LLM-Assisted Translation of Legacy FORTRAN Codes to C++: A Cross-Platform Study Proceedings of the 1st Workshop on AI and Scientific Discovery, NAACL 2025.

Contact this candidate