Kaustubh
Patil

Research Intern @ TUM | MIT-WPU CS'26

Core ML RL Fuzzing
// OPEN FOR WORK // CS'26 // ML RESEARCHER // PYTHON // LLMs // RAG // PYTORCH // DOCKER // OPEN FOR WORK // CS'26 // ML RESEARCHER // PYTHON // LLMs // RAG // PYTORCH // DOCKER // OPEN FOR WORK // CS'26 // ML RESEARCHER // PYTHON // LLMs // RAG // PYTORCH // DOCKER //
XP

Experience

TUM - Technical University of Munich

Research Intern | Nov 2025 - Present

Diffusion Sampling
  • Researching Diffusion Sampling in Reinforcement Learning contexts.
  • Investigating diffusion probabilistic models to enhance exploration strategies for RL agents.
  • Implementing novel sampling techniques to significantly improve sample efficiency in high-dimensional continuous control tasks.

UIUC - University of Illinois Urbana Champaign

Research Intern | May 2025 - Oct 2025

Fuzzer Optimization
  • Selected for competitive program. Optimizing fuzzers for code coverage.
  • Modified AFLGo's distance calculator for 49.5K+ code blocks.
  • Reduced computation time by 95.53% via precomputed graphs.

IIT Hyderabad

Research Intern | Jul 2025 – Aug 2025

Diabetic Retinopathy
  • Processed 300+ retinal fundus images using ViT (Vision Transformers).
  • Obtained 92% classification accuracy and improved inference speed by 17%.
  • Leveraged LIME and GradCAM for explainable AI (xAI).
  • Generated heatmaps highlighting retinal regions with 84% alignment to ophthalmologist annotations.

Leadership

Vice President

CSI MIT-WPU Student Chapter | Jun 2024-25

  • Organized and managed coding contests, hackathons, ideathons, and public speaking events.
  • Supervised internal club initiatives, facilitating a Research & Innovation Wing.

Research Mentor

CSI MIT-WPU | 2024-Present

  • Mentored over 10 students across various projects related to diverse ML topics.
  • Resulted in 2 papers accepted at conferences.

Tech Head

HackMIT-WPU | Feb-Mar 2025

  • Supervised technical tracks for 300+ teams and 1000+ participants.
  • Successfully organized the HackAIthon track.
  • Coordinated with developers/juries for smooth evaluations.
★
★

Achievements

Best Paper Award at ICCUBEA 2025 conference.

Certificate of Merit: Machine Learning camp, National Immersion Program at NIT Goa.

Consolation Prize at 14th Project Innovation Contest 2025 (21st ICDCIT).

Student Achiever Award in recognition of my achievements in 2024-25.

Outstanding Contribution Award 2024-25 by CSI MIT-WPU.

Rank Holder Scholarship.

Merit Scholarship

Runner-Up at HackMITWPU '24 Ideathon (Mental Health App).

Runner-Up in Bytes4Charity Hackathon (NGO Website).

Projects

TensorFlow

Sleep Stage Classification

CNN + BiLSTM

Multi-modal deep learning model achieving 93.60% accuracy for sleep stage classification.

91.89% Apnea Detection GITHUB >
DeepSeek

MailMind AI Agent

Python | Deepseek LLM

Autonomous multi-agent system reducing email processing time. Includes HITL.

98% Accuracy GITHUB >
Gemini

Breeze CLI Tool

Python | AST

Personalized CLI agent for code analysis. 12 polymorphic agent nodes with improved error handling.

20+ Languages GITHUB >

RESEARCH

Comparative Analysis of Deep Learning Models for Underwater Trash Detection: DeepLabv3+, R-CNN, YOLO, SSD

Springer

Intl Conf. on Intelligent Systems and Applications 2025.

First Author.

Dermatological Diagnosis w/ Few-Shot Learning

IEEE

Intl Conf. on Computing, Communication, Control & Automation 2025.

First Author.

Blockchain Degree Authentication

IEEE

Intl Conf. on Progressive Innovations in Intelligent Systems 2024.

Preserving Data Integrity and ACID Properties.

Plant Disease Detection (Corn, Banana, Rice)

Springer

Intl Conf. on Intelligent Systems and Applications 2024.

Using CNN.

ARSENAL
+
+

Technical Skillset

PYTHON RAG TENSORFLOW PYTORCH DOCKER SQL LANGCHAIN AI AGENTS
LET'S TALK

EMAIL ME

[kaus77135@gmail.com]
LINKEDIN GITHUB