Elice Priyadarshini

Published ML Researcher | 2 Peer-Reviewed Papers | ICML 2025 Contributor

Published ML researcher and engineer with 2 peer-reviewed publications and 4 years production experience. First-author publication in Springer (2024). Co-author in Journal of Integrated Science and Technology (2024). Engineering contributor to VersaPRM (ICML 2025).

At Siemens (2020-2024), led GPT-3.5 fine-tuning for sustainability recommendations, demonstrated to CEO and deployed to enterprise production. Built secure TAPAS-based chatbot with role-based access control. Shipped 17+ production applications serving thousands of users.

Master's student at UW-Madison (graduating Dec 2025). Actively seeking ML Engineer roles in LLM systems, evaluation infrastructure, or ML platforms.

Get In Touch GitHub Resume

Hi! I'm Elice :)

Publications

Analysis of Sea Surface Temperature and Chlorophyll-a Concentration Along the Coastline of the Indian Peninsula Using Remote Sensing Data

E. Priyadarshini, A.M. Deshpande, A. More, S. Pate
Springer - Geo-Environmental Hazards using AI-enabled Geospatial Techniques, 2024

A new framework for accessing and visualizing Ocean Color Data for water quality parameter analysis

A. Deshpande, E. Priyadarshini, A. More, S. Pate
Journal of Integrated Science and Technology, 12(1), 707-707, 2024

Key Achievements

Published Research

First-author Springer publication (2024). Co-author in peer-reviewed journal (2 citations). Research on remote sensing for water quality analysis using ML techniques.

ICML 2025 Contribution

Engineering contributor to VersaPRM - first multi-domain Process Reward Model. Built evaluation infrastructure achieving 7.9% improvement in non-math reasoning.

Production LLM Systems

Led GPT-3.5 fine-tuning at Siemens, deployed to CEO and enterprise customers. Built secure TAPAS chatbot with role-based access control for enterprise databases.

Experience

SQLPilot.ai - ML Engineer (May 2025 - Aug 2025)

Built RAG pipeline for Text-to-SQL system with 18% schema adherence improvement. Deployed production inference API with FastAPI, implemented A/B testing workflows and Docker containerization.

Lee Labs, UW-Madison - Graduate Researcher (Sep 2024 - Dec 2024)

Engineering contributor to VersaPRM (ICML 2025)

Built evaluation infrastructure using AWS SageMaker for large-scale PRM benchmarking, accelerating evaluation by 40% across 10+ datasets. Conducted ablation studies on inference algorithms (Beam Search, MCTS, Best-of-N). Contributed to open-source release.

Siemens Digital Industries - Software Engineer (Sep 2020 - Jul 2024)

ML/AI Leadership: Led GPT-3.5 fine-tuning for BOM sustainability, deployed to production Teamcenter PLM, demonstrated to CEO. Built secure TAPAS-based chatbot with role-based access control for enterprise database queries.

Research: Published 2 peer-reviewed papers on remote sensing and ML for water quality analysis.

Platform Engineering: Authored API design whitepaper adopted by 17+ teams. Built shared component library reducing development time by 30%. Mentored 4 junior engineers.

Early Experience (2019-2020)

TATA AutoComp: Built YOLOv3-based pedestrian detection and driver attention monitoring for ADAS systems.

AI & CV Society: Co-founded technical society, designed ML workshops, mentored students, published tutorials.

Featured Projects

VersaPRM

ICML 2025 Research

Engineering contributor to multi-domain Process Reward Model research. Built evaluation infrastructure using AWS SageMaker, conducted ablation studies demonstrating 7.9% accuracy gains.

PyTorch LoRA AWS SageMaker MCTS

Remote Sensing Research

2 Published Papers

First-author Springer publication on sea surface temperature and chlorophyll analysis. Co-author journal paper on ocean color data framework. 3 total citations.

Remote Sensing MODIS Python

SQLPilot.ai

Production

Text-to-SQL system with RAG achieving 18% schema adherence improvement. Production API with A/B testing and Docker deployment.

LangChain FastAPI Docker

Siemens ML Systems

Enterprise Production

GPT-3.5 fine-tuning for sustainability (CEO demo). TAPAS chatbot with secure database access. 17+ shipped products.

GPT-3.5 TAPAS OpenLCA

Wafer Defect Detection

Computer Vision

CNN/ViT classifier achieving 96% F1-score for nanoscale defects. 45% inference speedup using TensorRT optimization.

PyTorch TensorRT ONNX

CUDA Optimization

HPC

CUDA kernel optimization achieving 300x speedup over CPU through shared memory and thread synchronization.

CUDA C++ Performance