Harshith Kethireddy MS AI · University of Michigan–Dearborn

Open to full-time ML roles

Harshith
Kethireddy

MS AI student at the University of Michigan (4.0 GPA) with hands-on research in preference-based reinforcement learning and two peer-reviewed publications. Building ML systems that actually work — at IQVIA, in the lab, and in production.

Get in Touch View Resume

4.0/4.0 MS GPA

2 Publications

3 Research Internships

🎤 3 Minute Thesis Competition at UMich Dearborn

About Me

Building ML that moves the needle.

Originally from India, I moved to the US in Fall 2024 to pursue my MS in AI at UMich Dearborn. I'm passionate about reinforcement learning, deep learning, and turning research into real-world impact. From benchmarking production ML pipelines at IQVIA to publishing at ICPR and VLSID, I bring a rare combination of academic rigor and engineering execution.

🔬
Research-First
Novel frameworks for preference-based RL with paper under review.

⚙️
Engineering Depth
End-to-end pipelines: SAC agents, ensemble reward models, KL selection.

📊
Industry Impact
+15 PR-AUC points on real pharmaceutical datasets at IQVIA.

🌐
Breadth
Computer vision, NLP, gradient boosting, contrastive learning.

Education

🟦

University of Michigan – Dearborn

M.S. Artificial Intelligence (Machine Learning)

Sep 2024 – May 2026 · Dearborn, MI

⭐ 4.00 / 4.00 GPA

Advanced Data Mining Deep Learning Text Mining & IR Master's Thesis

🟧

IIT Jodhpur

B.Tech. in AI & Data Science Engineering

Dec 2020 – May 2024 · Jodhpur, India

7.17 / 10.00 GPA

Computer Vision Dependable AI Pattern Recognition

Where I've Worked

Experience

IQVIA · King of Prussia, PA

Machine Learning Innovation Intern

📍 On-site

Jun 2025 – Sep 2025

XGBoost LightGBM Imbalanced Learning Pharmaceutical Analytics PR-AUC Optimization

Benchmarked classification algorithms (XGBoost, LightGBM, RF, logistic regression) to identify cardiology- and dermatology-associated Advanced Practice Providers under strict compute constraints.

Improved Precision–Recall AUC from 0.67 to 0.82 via strategic downsampling, class-weighted loss, and threshold tuning on imbalanced pharmaceutical-analytics datasets.

Curated and validated large-scale 2025 APP datasets, cutting label noise for downstream ML pipelines used across therapeutic-area models.

IIT Jodhpur · CS Department · In collaboration with Swiggy

Summer Research Intern

📍 Jodhpur, India

May 2023 – Jul 2023

Face Recognition FSCIL Continual Learning CiFAR-100

Developed face recognition software using feature mapping to identify individuals from selfies; tested for accuracy and security robustness.

Addressed catastrophic forgetting using Constrained Few-Shot Class-Incremental Learning (FSCIL) on CiFAR-100 and trained the D-LORD dataset for facial verification.

Advisor: Dr. Richa Singh

Robert Bosch Centre for Cyber-Physical Systems · IISc Bangalore

Summer Research Intern

📍 Bangalore, India

May 2022 – Jun 2022

Simulation Leaflet.js Mobility Modeling

Built a marker simulation of multimodal transport flows (pedestrians, vehicles) moving across a city map using the Leaflet.js geospatial library.

Advisors: Dr. Rajesh Sundaresan, Mr. Jagadish Midthala

Academic Coursework

Courses

All coursework from my graduate studies at UMich Dearborn and undergraduate at IIT Jodhpur.

Semester I · Aug – Dec 2024

CIS 579 Artificial Intelligence

CIS 553 Software Engineering

CIS 350 Data Structures & Algorithm Analysis

Semester II · Jan – Apr 2025

CIS 5xx Text Mining & Information Retrieval

CIS 5700 Advanced Data Mining

ECE 579 Intelligent Systems

Semester III · Sep – Dec 2025

CIS 5xx Algorithm Analysis & Design

CIS 583 Deep Learning

CIS 699 Master's Thesis

Semester IV · Jan – Apr 2026

CIS 5xx Advanced Artificial Intelligence

CIS 699 Master's Thesis

Semester IV · Jan – Apr 2022

CSL2050 Pattern Recognition & Machine Learning

CSL2080 Principles of Computer Systems 1

CSL2090 Principles of Computer Systems 2

CSP2020 Human Machine Interaction

Semester V · Jul – Dec 2022

CSL4010 Optimization for Machine Learning

CSL3090 Artificial Intelligence

CSL3010 Design & Analysis of Algorithms

CSL4030 Data Engineering

Semester VI · Jan – May 2023

CSL4020 Deep Learning

CSL7580 Advanced Artificial Intelligence

CSL7370 Dependable Artificial Intelligence

CSL4050 Data Visualization

CSL7650 Autonomous Systems

Semester VII · Jul – Dec 2023

CSL7420 Principles of Biological Vision

AIL7250 Computational Cognitive Neuroscience

CSL7490 Introduction to Blockchain

CSD4030 Tech Project

Semester VIII · Jan – May 2024

CSL7360 Computer Vision

EEL3100 Bioimaging

CSL7080 Security & Applications

MAL4330 Intro to Financial Engineering

Academic Work

Research

Advancing preference-based RL and applied deep learning, with work published at top venues.

Preference-based RL Under Review

Trajectory Feature Magnitude Noise for Preference Uncertainty Modeling

Designed a novel framework modeling preference uncertainty from torque instability in continuous-control tasks (DMControl, Meta-World). Extends PEBBLE with latent-embedding reward learning: frozen InfoNCE encoder + sigmoid-weighted Bradley–Terry loss across three distinguishability-query variants. Built an end-to-end RL pipeline with SAC agent, ensemble reward models, and KL-based trust-sample selection — scaling to multi-seed benchmarking across 4 domains.

SAC InfoNCE Bradley–Terry DMControl Meta-World

🏛 University of Michigan – Dearborn · Advisor: Dr. Srijita Das · Nov 2024 – Present

🧮 Working through preference-based RL theory

Publications

Peer-Reviewed Papers

ICPR 2024 · Springer

Fusion Autoencoder for Differential Diagnosis of Thyroid Tumors

Harshith Kethireddy, Dr. A. Paul

Read Paper

VLSID 2025 · IEEE · ORAL

Meta-Training for Multi-Object Detection on Resource-Constrained UAVs

Harshith Kethireddy, Dr. B. Kumar

Read Paper

Selected Work

Projects

🌲

Knowledge-Intensive Gradient Boosting (KiGB)

Paper Reimplementation

Reimplemented soft-monotonicity constraints in gradient boosting; benchmarked against LightGBM, XGBoost, Monoensemble across 15+ UCI datasets and 2 real-world datasets. Studied λ/ε sensitivity and conflicting-advice regimes. Outlined extensions and characterized data-scarcity behavior.

Gradient Boosting LightGBM XGBoost Python

View on GitHub

✋

Real-Time Hand Gesture Recognition

Custom CNN + Transfer Learning

Fine-tuned MobileNetV2 on a self-collected 13,200-image, 11-class dataset; deployed a real-time MediaPipe hand-tracking demo. Comparative study of MobileNetV2, ResNet50, and a 13-layer custom CNN covering preprocessing, training, and hyperparameter-tuning pipelines.

⚡ 98.8% test accuracy

MobileNetV2 PyTorch MediaPipe Transfer Learning

View on GitHub

Beyond the Lab

Extracurricular Activities

Leadership, mentorship, and community involvement at IIT Jodhpur.

📍 Times Square, NYC

🎓

Class Representative

AI & Data Science Engineering · Class of 2024

Indian Institute of Technology Jodhpur

Served as a vital link between students and faculty — gathered feedback and represented student concerns to improve the learning experience.

Organized meetings to discuss course-related issues and facilitated discussions on curriculum improvements.

Outcome: Enhanced leadership and communication skills supporting 60+ classmates throughout the program.

🎵

Mentor

Sangam — Music Club

Indian Institute of Technology Jodhpur

Guided members in musical pursuits, fostering collaboration and skill development across diverse instruments and styles.

Helped organize workshops and jam sessions that facilitated collaboration among musicians.

Outcome: Developed inclusive leadership — building an environment where every member felt supported to grow.

📚

Mentor

Lit Soc — Literature Club

Indian Institute of Technology Jodhpur

Organized poetry readings, book discussions, and creative writing events encouraging critical thinking and expression.

Provided constructive feedback and facilitated engaging discussions, helping peers articulate ideas clearly.

Outcome: Strengthened communication and analytical skills through exposure to diverse literary forms.

📣

Public Relations Team Assistant Head

Varchas'22 — Annual Sports Fest

Indian Institute of Technology Jodhpur

Crafted and spread key messages about Varchas'22 across platforms; led a team of volunteers and coordinated with campus and local media.

Managed the fest's social media presence to maximize outreach and participant engagement.

Outcome: Learned to handle unexpected situations under pressure and maintain a positive public image during a large-scale event.

🚌

Accommodation & Transport Team Assistant Head

Varchas'22 — Annual Sports Fest

Indian Institute of Technology Jodhpur

Organized and managed transportation schedules for participants, ensuring timely arrivals and departures.

Demonstrated quick thinking and adaptability when resolving unexpected issues like last-minute changes.

Outcome: Strengthened vendor and stakeholder communication, and built resilience managing logistics at scale.

⚙️

Tech Events Core Member

Prometeo'22 — Annual Tech Fest

Indian Institute of Technology Jodhpur

Brainstormed and developed engaging technical events — competitions and workshops appealing to students across disciplines.

Contributed to introducing new tech-based competitions, fostering innovation and cross-team collaboration.

Outcome: Gained experience managing multiple tasks simultaneously and meeting tight deadlines in a fast-paced environment.

Technical Toolkit

Skills

Programming Languages

Python95%

C++75%

SQL80%

Bash / Shell70%

ML / DL Frameworks

PyTorch92%

TensorFlow / Keras80%

Scikit-learn90%

Hugging Face78%

Tools & Infra

Git / Docker / Linux88%

Weights & Biases82%

Focus Areas & Specializations

Reinforcement Learning Deep Learning PyTorch Preference-based RL Contrastive Learning Meta-Learning Transfer Learning CNNs Transformers Uncertainty Estimation SAC Agent PEBBLE InfoNCE Bradley–Terry Gradient Boosting XGBoost LightGBM Computer Vision NLP Imbalanced Learning Ensemble Methods DMControl Meta-World MediaPipe

Currently Exploring

RLHF LLM Fine-tuning Reward Modeling Multi-agent RL Foundation Models