Krishnanand Yadav

I’m a hands-on developer and research intern who turns real-world problems into working systems—pairing machine learning fundamentals with production-minded software engineering. I enjoy shipping features end-to-end, from data/ML design through scalable APIs and responsive user experiences.

In my summer research internship, I designed and trained a Convolutional Autoencoder in Python (TensorFlow, Keras) for multi-class colonoscopy image classification on a 5000+ image dataset. I also improved training pipelines to increase iteration speed and reduce training instability across experiments.

On my LLM-assisted cybersecurity project, I built an LLM + RAG rule generator grounded on real attack data, achieving 100% syntactic validity and 87.5% detection across representative attack scenarios in a pfSense + Suricata testbed. I also trained XGBoost classifiers on UNSW-NB15 using a custom feature vector (AUC-ROC of 0.9795), and engineered a streaming workflow to classify live network traffic.

I balance performance and correctness in application engineering too. I built a real-time collaborative whiteboard with shared canvas state, live cursors, RBAC + authentication (OTP + OAuth), and low-latency presence using Supabase Edge Functions and Liveblocks, while reducing frontend resource size by 72% and improving LCP from 4.5s to 2.1s.