Research

Making AI systems reliable enough to trust.

I work on applied problems where mistakes are expensive: industrial systems, public services, elections and homes.

Publications

  • 2025

    Automated Invoice Data Extraction: Using LLM and OCR

    K. Khanchandani, A. Thakur, A. Shetty, C. Reddy, R. Behera

    arXiv:2511.05547

    Read on arXiv
  • 2025

    An Accessibility-First Framework for Enhancing Web Experiences of People with Disabilities

    A. Thakur, H. Harianawala, P. Trivedi, D. Patole

    IEEE ICCUBEA 2025 · peer-reviewed

    Read on IEEE Xplore

Current interests

  1. 01Edge AI that runs on the device, not the cloud
  2. 02Machine intelligence for industrial telemetry and control
  3. 03Hallucination-resistant document understanding
  4. 04Local-first, low-latency voice AI
  5. 05Verifiable systems for civic trust
  6. 06Security and governance for connected IoT devices

Contact

Let's build something.

Projects, research, speaking or just a good problem — my inbox is open.

Email meMore ways to reach me