If It's Broken,Let's Fix It With AI
Some people see problems. I see systems—ones that could be smarter, faster, and more efficient. Some might call it obsession; I call it curiosity that refuses to settle.
It starts with a simple question: Why isn't this working better? That question leads to deep dives into data, endless iterations of models, and solutions that don't just fix problems—they redefine what's possible. It's the reason I spend hours refining AI-driven optimizations, not because I have to, but because stopping at "good enough" isn't really an option.
The Work Speaks for Itself
When something doesn't perform the way it should, I don't accept limitations—I reengineer them. Whether it's making AI understand complex patterns or refining predictions until they're as close to reality as possible, every challenge is a new equation waiting to be solved.
How I Approach Problems
- Data is the blueprint – AI is only as good as the insights it learns from.
- No assumptions, only testing – Iteration is key to fine-tuning models.
- AI should be practical – Tech isn't valuable unless it creates real change.
- Innovation never stops – If something can be improved, it should be.
Latest Projects
View All →Projects & Experiments
AI-Powered Traffic Management
Professional Journey
A timeline of my academic and professional experiences, showcasing my growth and achievements in AI and technology.
Education
- Specialized in Artificial Intelligence and Machine Learning
- Graduated with honors (9.2/10 GPA)
- Relevant coursework: Machine Learning, Deep Learning, Computer Vision, Natural Language Processing, Data Structures & Algorithms
Internships
AI Research Intern
Google Research- Worked on improving object detection models for low-light conditions
- Implemented and fine-tuned a novel attention mechanism that improved accuracy by 15%
- Contributed to an internal research paper published in the company's technical journal
Machine Learning Engineer Intern
Microsoft- Developed natural language processing models to improve text summarization in Microsoft Edge
- Achieved 22% improvement in ROUGE-L score compared to the previous model
- Worked with a team of engineers to deploy the model to production
Research
AI-Powered Smart Traffic Management System
IIT Kharagpur Research Lab- Designed and implemented a computer vision-based traffic monitoring system
- Created algorithms to dynamically optimize traffic signal timing based on vehicle density
- Reduced average waiting times at intersections by 27% in simulation
- Published findings at the International Conference on Smart Cities
AI-Optimized Drug Solubility & Bioavailability Prediction
Interdisciplinary Research Program- Collaborated with the Department of Pharmaceutical Sciences to develop AI models for predicting drug properties
- Built a graph neural network to predict solubility and bioavailability of novel compounds
- Achieved 91% accuracy in predicting solubility classes, outperforming traditional QSAR methods
- Model is being used to screen potential drug candidates and accelerate development
Open Source
Contributor
Hugging Face Transformers- Regular contributor to the Hugging Face Transformers library
- Implemented optimizations for T5 model inference that improved speed by 12%
- Added documentation and examples for using transformers in low-resource settings
Hackathons
First Place - AI for Healthcare Hackathon
MIT Hacking Medicine- Developed an AI system to detect early signs of diabetic retinopathy from retinal scans
- Achieved 94% sensitivity and 92% specificity on the test dataset
- Created a web application for doctors to upload and analyze patient scans
- Won first place among 120+ teams
Leadership
President
AI & ML Club, IIT Kharagpur- Led a student organization of 200+ members interested in AI and ML
- Organized workshops, hackathons, and guest lectures from industry experts
- Mentored 30+ students on AI/ML projects and career paths
- Established partnerships with 5 tech companies for sponsorships and internship opportunities
Certifications
AI & ML Professional Certification
IIT Kharagpur & Intellipaat- Comprehensive certification covering machine learning, deep learning, and AI applications
- Completed 20+ hands-on projects across computer vision, NLP, and reinforcement learning
- Received distinction for capstone project on multi-modal sentiment analysis
Deep Learning Specialization
Coursera (Andrew Ng)- Completed 5-course specialization covering neural networks, CNN, RNN, and more
- Built and trained deep learning models for various applications
- Implemented research papers and state-of-the-art architectures
Skill Distribution
Core Expertise
Key Skills
Tech Stack
Research Publications
Exploring the frontiers of artificial intelligence through peer-reviewed research and academic contributions
Attention-Based Multi-Modal Fusion for Enhanced Medical Image Diagnosis
We propose a novel attention mechanism for fusing multi-modal medical imaging data, demonstrating significant improvements in diagnostic accuracy across multiple conditions. Our approach leverages transformer architectures to dynamically weight different imaging modalities.
Self-Supervised Learning for Real-Time Traffic Flow Prediction
A novel self-supervised learning approach for traffic flow prediction that leverages unlabeled traffic camera data. Our method achieves state-of-the-art performance while reducing the need for expensive labeled data.
Efficient Graph Neural Networks for Large-Scale Molecular Property Prediction
We introduce a scalable graph neural network architecture for molecular property prediction that achieves linear complexity while maintaining high accuracy. The method is validated on multiple pharmaceutical datasets.
Get In Touch
Let's create something amazing together
© 2025 Akhil Pavuluri. All rights reserved.