rockh3x

Vivek Patel

Ph.D. researcher building machine learning that runs on drones, edge devices and GPUs — under real time budgets.

See the research View the code
Scroll

Ten thousand points of data.

Raw sensor input arrives unstructured, noisy, and far larger than the device that has to process it.

Structure, found under constraint.

Models learn the shape hidden inside it — while fitting into memory, power and latency budgets that leave no room to spare.

Decisions, in real time.

What comes out the other side is small enough, fast enough, and reliable enough to run where it actually matters.

Focus

Four problems.
One underlying question.

How much intelligence can you fit into a device that has almost nothing to spare?

Edge AI

Detection frameworks designed for drone-mounted deployment, where compute, memory and power are all tightly bounded.

UAV

GPU acceleration

Moving search and recovery workloads off the CPU and onto thousands of CUDA cores.

CUDA

Computer vision

Adaptive enhancement for imagery that breaks ordinary pipelines — underwater colour attenuation, turbidity, aerial scale.

Vision

Systems & protocols

Semantic tokenization and deterministic transport, plus earlier work on blockchain wallet infrastructure.

DCST
By the numbers

A record, still being written.

0
Research projects
Public repositories
0
Research areas
0
Platforms online
Research

Selected work.

Five projects across vision, acceleration, transport and cryptography.

Software

Public repositories.

Pulled live from GitHub each time this page loads.

Loading…

Connect

Elsewhere.