Pleasanton, CA
Basil Khwaja
Rockstar Researcher
Learningtolivelife.
Teachingmachinestolivelikeonetoo.
Me.
Hello! I'm Basil Khwaja, an aspiring AI/ML Engineer. I'm currently a Master's student at Purdue University, pursuing a degree in Electrical and Computer Engineering. I am also a Research Fellow at Purdue. My experience includes working at a startup and mentoring students on the importance of engineering problem-solving.
Born and raised in Bangalore, India. I moved to America at the age of seven. However, my interests extend beyond technology—I am also an aspiring dancer, currently exploring contemporary and hip-hop styles. In addition, I'm learning to play the guitar and part of a band Holladaze! Recently, I've taken up bouldering, and I'm excited to see where that journey takes me.
Education:
- Bachelor's Degree in Computer Engineering, Purdue University (2022 - 2025)
- Master's Degree in Electrical and Computer Engineering, Purdue University (2025 - 2026)
Publications.
Thesis
Master's ThesisPurdue University · 2026
AEM : Adaptive Edge MPC via Non Stationary Mixture Bandits
Basil Hasan Khwaja
Classical Model Predictive Control (MPC) for autonomous vehicles typically samples trajectories uniformly from a fixed motion-primitive library, which is computationally inefficient and fails to leverage knowledge from past driving experiences. We propose Adaptive E-MPC (AEM), which replaces uniform sampling with an adaptive mixture distribution that balances exploration and exploitation using prior trajectory performance. To avoid the computational expense of repeatedly searching for the optimal mixing parameter, AEM combines a confidence-aware online heuristic with an Adaptive NeuralUCB contextual bandit that learns residual corrections and accounts for uncertainty in changing environments. Experiments in autonomous driving simulations show that AEM reduces cumulative driving cost compared to uniform sampling, while the bandit-enhanced version provides additional performance gains with bounded regret guarantees.
Computer EngineeringAutonomous VehiclesModel Predictive ControlContextual BanditsTrajectory Planning
Projects.
From Me To YouA passion project I have been working on for the past few months is a web app that allows users to send a message to a loved one and receive a thoughtful response in return. It is built using Next.js, Tailwind CSS, Flask, Firebase, and Typesense, giving me the chance to work across the full stack of web development. The app is hosted on Vercel, and the server is deployed through Heroku. This project has helped me deepen my understanding of modern web architecture while creating something meaningful and emotionally engaging. More updates coming soon!
Swerve (HackTech - Caltech)
Hackathon Winner
- Designed and developed Hugo, an AI-powered procurement assistant using LangChain, GPT-4o, and GPT-3.5 Turbo to automate inventory forecasting, supplier communication analysis, and strategic purchasing decisions.
- Integrated ERP, CAD, and unstructured data sources into actionable insights through a custom React/Tailwind dashboard, enhancing decision-making for procurement teams in complex supply chains.
NoSu (HackMIT - MIT)
- Built a full-stack system that analyzes short videos with computer vision (YOLOv5, BLIP, VideoMAE) and audio sentiment analysis, then generates and syncs custom background music using LLMs + Suno API
- Integrated FastAPI backend, React/Tailwind frontend, and Firebase (Auth, Firestore, Storage) to manage video uploads, AI analysis outputs, music generation, and automatic video/audio muxing.
MuSHR MPC Integration
- Designed complete path planning and trajectory generation pipeline
- Developed process for environmental data collection and map generation
- Created and implemented three unique path-planning algorithms
- Integrated algorithms with cubic spline-based trajectory method
Computer Vision and Generative Models
- Implemented CMMD to measure alignment between conditional distributions
- Investigated methods of detecting correct logic between prompt and model generation through image detection (Yolo, R-CNN, and DETR)
Teach Me (AI Atlanta Hackathon)
- Developed a platform to use all the LLM models (Gemini and Claude-Sonnet) in one place to enhance learning experience
- Teach Me bridges the gap from diverse dynamic learning information to a catered learning style
SWE-agent reimplementation
- Built on top of large language models, it can understand GitHub issues, navigate complex codebases, and generate meaningful code changes
- Understanding a task, planning, editing code, running tests, and even making pull requests
- Streamline the development process and explore the future of autonomous software engineering
Trace AI (HackGT - GeorgiaTech)
- Developed an algorithm to give credibility to creators and choreographers for the work they put out on social media
- Utilizing Tensorflow and Panda3d, we modeled human movements and stored them in the database
- Assigning credibility to artists that made the original creation while not having to worry about credibility be lost
AI-Blocks
- A comprehensive collection of core components for understanding and building AI systems from the ground up
- Module is designed to be beginner-friendly, yet powerful enough for experimentation and extension
Skills.
Python
C
C++
Java
React
Pytorch
Tensorflow
ROS
Panda3d
Three.js
Lang Chain
Spring Boot
Music.
Juxtaposed
Lost Keys
Suffocate
Space
Quotes I love
"Medicine, law, business, engineering—these are noble pursuits and necessary to sustain life, but poetry, beauty, romance, love—these are what we stay alive for."- Dead Poet's Society
"But how could you live and have no story to tell?"- White Nights
"One could not count the moons that shimmer on her roofs, or the thousand splendid suns that hide behind her walls"- A Thousand Splendid Suns