ML engineer for multichannel audio. I build acoustic detection systems on microphone arrays: synthetic data pipelines, deep learning, and GPU signal processing on NVIDIA Jetson. I'm also co-founder and CTO of The Algorithm Lab, an early-stage AI startup.
Portfolio 路 Google Scholar 路 LinkedIn 路 arukhandelwal281@gmail.com
- Earmark, built in public: personal voice isolation and barge-in gating for voice agents, running in the browser. After a 5-second enrolment, a causal model keeps only your voice and tells the agent when you, and only you, are speaking. PyTorch for the model and a C++17 engine for WebAssembly. Work in progress: no results until they come from real runs.
- Audio ML at SSS Defence: acoustic drone detection on a multi-microphone array. The work spans synthetic data generation, sim-to-real transfer, CUDA signal processing and deployment on NVIDIA Jetson. The code is private.
- The Algorithm Lab: the product behind our AI marketing service, now in beta. A multi-agent system (platform writers, a topic planner, an adversarial critic and a campaign planner) working from one shared company audit, plus a website-audit pipeline, on FastAPI and React with several models through OpenRouter. The code is private.
- Unsupervised Approach for Classifying Volume Scattering for Surface Roughness and Potential Water-Ice Deposits. IEEE InGARSS 2025. IEEE Xplore
- Beyond Nonce Uniqueness: A Cryptographic Evaluation of AES-GCM in Secure Anonymous File Sharing Systems. ICDSA 2025, first author. Springer
| Project | What it is | Stack |
|---|---|---|
| Earmark | Personal voice isolation and barge-in gating for voice agents, in the browser (work in progress) | PyTorch, C++17, WebAssembly |
| Hypercharge | Wires a repo for Cursor and Claude Code agents: a code map, session memory and safety hooks | Python, tree-sitter, SQLite |
| SwiftShare | Anonymous, encrypted file sharing; the basis of my ICDSA paper | FastAPI, AES-GCM, AWS S3 |
| KALI | NSE paper-trading sandbox: 4 algorithm ledgers, HMM regime features, Kelly sizing | Python |
| Review Insights | LLM review analytics with Google Business Profile sync | Next.js, PostgreSQL, Prisma |
| Sensor Anomaly | Real-time anomaly detection on NASA turbofan sensor data | scikit-learn, FastAPI |
Python 路 PyTorch 路 C++ 路 CUDA 路 NVIDIA Jetson 路 WebAssembly 路 FastAPI 路 PostgreSQL 路 TypeScript 路 React 路 Next.js 路 AWS