Skip to content
View kpatherya's full-sized avatar
  • Atlanta, GA

Block or report kpatherya

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
kpatherya/README.md

Kausar Patherya

CS PhD researcher at Georgia Tech, building efficient, inspectable AI systems over multimodal data. Open to collaboration on research software, evaluation, and AI systems that need to be reproducible and inspectable.

Projects

Flash-Fusion Natural-language queries over IoT sensor streams, compiled into typed plans and executed deterministically. 📉 95% lower latency · 98% lower token cost vs. feeding raw telemetry to an LLM. arXiv Demo

POLARIS Vision-language pipeline matching landmarks across seasonal change — OWL-ViT → FastVLM → depth → RANSAC.

Sting-Sense Bus-fleet telemetry system: Zephyr firmware on nRF9160 (bus-rtos) with route congestion analytics. Demo

SIT-UP Camera-based posture sensing with intervention prototypes.

Reach out

🌐 kausarpatherya.com — write-ups and project pages 🎙️ Georgia Tech AI Podcast — host and producer

📧 Email · 📄 CV · 🎓 Scholar · 💼 LinkedIn

Pinned Loading

  1. flash-fusion flash-fusion Public

    Expressive, low-latency natural-language queries over IoT sensor streams: typed analysis plans, schema validation, deterministic operators (arXiv:2511.11885)

    Jupyter Notebook

  2. sting-sense sting-sense Public

    Interactive analytics for Georgia Tech bus-route telemetry: traffic scoring, hour filtering, and route-level congestion maps

    HTML

  3. polaris polaris Public

    POLARIS: staged vision-language pipeline for cross-temporal landmark matching with OWL-ViT detection, FastVLM enrichment, and geometric verification

    Python

  4. situp situp Public

    SIT-UP: real-time posture sensing from multi-camera video with audio, vibration, and desk-actuation intervention prototypes

    Python