Mahendra Avudiyappan
AI / ML Engineer
AI / ML Engineer, helping organizations unlock insights from data using machine learning and AI.
AI / ML Engineer
AI / ML Engineer, helping organizations unlock insights from data using machine learning and AI.
I’m an AI / ML Engineer based in Chicago who enjoys building practical machine learning solutions that help organizations make better decisions with data. My work focuses on applied machine learning, where I take real-world datasets and turn them into reliable, interpretable models. I have hands-on experience designing end-to-end ML pipelines, including data cleaning, feature engineering, model training, evaluation, and interpretation. A key part of my work has been building anomaly detection systems for healthcare workflows, where I used deep learning models to identify delays, inconsistencies, and operational patterns in complex processes.
I also work with GenAI and LLM-based applications, evaluating open-source language models and building Retrieval-Augmented Generation (RAG) systems to produce more accurate and context-aware outputs. I’m currently seeking entry-level AI / ML Engineer roles where I can continue learning, collaborate with experienced teams, and contribute to building data-driven, real-world AI systems.
Led applied machine learning initiatives to improve data reliability and support real-time analytics. Standardized and validated large-scale healthcare workflow data using automated checks. Designed anomaly detection and feature engineering pipelines with Python and PyTorch to improve predictive accuracy. Built and evaluated LLM and GenAI solutions, including RAG-based workflows, and streamlined experimentation using MLflow and containerization.
Optimized data preprocessing pipelines across 115K+ healthcare records
Improved data reliability using automated validation and drift checks
Enhanced anomaly detection using PyTorch MLP with focal loss
Improved anomaly precision through class weighting and feature design
Temporal, workload, insurance, and rolling-window features
↑ 11% throughput and ↓ 18% verification queue time
Improved response accuracy and ↓ 50% latency in GenAI applications
Applied ML and time-series modeling for budget planning systems
Built an end-to-end machine learning system to detect delayed and incomplete prescription workflows using 115K+ real-world healthcare records. Improved macro-F1 by 8.9 points and reduced false positives by 22% using a PyTorch MLP with focal loss.
Designed and evaluated LLM-based recommendation systems using LLaMA and DeepSeek models. Integrated Retrieval-Augmented Generation (RAG) to improve response accuracy and reduce latency by 50%.
Developed CNN and UNet-based deep learning models for MRI image classification and segmentation, achieving 97.36% accuracy. Published results in a peer-reviewed journal.
Built supervised ML and time-series models to improve budget forecasting accuracy by up to 20%, supporting data-driven financial planning for $100K+ budgets.
Python
SQL
JavaScript
HTML5
CSS
Git
GitHub
I’m always open to collaborating on impactful AI and machine learning projects or exploring new opportunities in applied data science. Whether you have a problem to solve or would like to connect, feel free to reach out—I’d be glad to chat.