Hi, I'm Devansh Khandelwal

Computer Science & Data Science Student

Pursuing BS in Computer Science and Data Science at Purdue University. Passionate about machine learning, software engineering, and building impactful solutions.

View My Work Get In Touch
Devansh Khandelwal

About Me

I'm a Computer Science student at Purdue University. My academic journey is complemented by minors in Mathematics and Behavioral Economics, giving me a unique perspective on problem-solving.

I'm passionate about machine learning, software engineering, and building systems that make a real impact. Currently working on climate prediction models and leading software development initiatives.

Technical Skills

Python Java C++ JavaScript PyTorch TensorFlow MERN Stack MongoDB PostgreSQL Git Linux CI/CD

Experience

SURF Researcher

Purdue Engineering - WCD Lab

May 2025 - Present
  • Building spatio-temporal CNNs on 165k+ climate samples for atmospheric blocking and heatwave prediction using multi-day inputs and physics-informed constraints.
  • Transferred CESM-LENS climate knowledge to ERA5 reanalysis, achieving ≥92% balanced TPR/TNR across lead times up to 12 days
  • Applied XAI (IXG, IG, GradCAM, LRP) to isolate anticyclone evidence for interpretable blocking-onset forecasts

Team Lead and Club Treasurer

Hack the Future

Aug 2024 - Present
  • Directing 10+ developers in a national non-profit software initiative, delivering a full-stack event platform with calendar sync & accessibility features.
  • Platform adopted by 200+ participants across biweekly events
  • Established CI/CD pipelines and automated testing, ensuring reliable deployments
  • Led weekly meetings mentoring teammates on Git, APIs, Agile and MERN stack development

ML Engineer

Purdue Vertically Integrated Projects - VIPER Lab

Aug 2024 - Dec 2024
  • Achieved 94% accuracy on 5k+ chess images using a neural network with automated square detection (R² = 0.984).
  • Implemented core CV algorithms (Canny, Otsu's Thresholding & Morphological Dilation) from scratch

Undergraduate Researcher

Cisco

Jan 2024 - May 2024
  • Engineered 2-stage hierarchical demand forecasting models, improving accuracy by 15%.
  • Collaborated with the Metrics and Evaluation team to define KPIs for production adoption

Featured Projects

Attention-Based SAC Portfolio Allocator

Developed SAC agent with multi-head attention, improving Sharpe ratio by +6.6% over equal-weight baselines. Implemented regime-aware rewards with interpretable attention, outperforming SPY-only by 30%+.

Python PyTorch Gymnasium yFinance Pandas
GitHub

NutriBuddy

Built a full-stack nutrition tracking platform, leveraging Llama 3.1 8B Instruct API to generate AI-powered recipes. Established personalized nutrition planning and pantry features based on Total Daily Energy Expenditure.

MongoDB Express.js React.js Node.js Material-UI Llama API
GitHub

Chess Position Recognition System

Achieved 94% accuracy on 5k+ chess images using a neural network with automated square detection (R² = 0.984). Implemented core CV algorithms from scratch including Canny edge detection, Otsu's Thresholding & Morphological Dilation.

Python PyTorch OpenCV Computer Vision
GitHub

Get In Touch

Let's Connect

I'm always interested in new opportunities and exciting projects. Feel free to reach out!

khanded@purdue.edu
765-543-7885
West Lafayette, IN