Engineer. I build.
Malicious URL detection using a probabilistic Bloom Filter pipeline. Ingests live threat feeds from URLhaus, PhishTank, OpenPhish, and Cert.pl.
AIS vessel anomaly detection — finds ships going dark, loitering, or moving erratically. Real-world use: sanctions evasion, illegal fishing, STS monitoring.
Physics-informed 1D CNN with CBAM attention that classifies radio signals into 11 modulation types from raw IQ data. Trained on RadioML 2016.10a.
Graph Attention Network for Twitter bot detection on cresci-2017. Statistically proves the model suppresses bot-to-human camouflage edges at p < 10⁻⁴¹.
Production-grade RAG system — upload any PDF, query it in plain English. Fully offline using Ollama. Custom RAGAS-style eval harness with experiment tracking.
Scrapes news articles and scores political bias from −50 (far-left) to +50 (far-right) using Gemini AI. Generates neutral summaries under 50 words.
eal-time multi-user chat application with WebSockets. React + Vite frontend, Flask + Socket.IO backend. One-command local startup, deployable to Render.
I'm Dhruv, a third-year Information Science student at BMS College of Engineering, Bangalore.
I'm interested in applied machine learning, systems architecture, and building software that holds up under
real-world constraints, not just inside notebooks.
Most of my work sits somewhere between
backend systems, probabilistic data structures, and machine learning.
I like problems that force you to understand what's actually happening under the hood.