Machine learning
Krishna
Models, experiments, perception, and the questions that turn raw sensor data into useful behavior.
A small engineering studio for curious prototypes
YantriKey
Three friends building things at the edge of robotics, machine learning, and software. First target: control an Elegoo tank robot from a phone over the internet.
Why this exists
It started as a reason to build together.
YantriKey is not a company pitch. It is a shared bench for three engineers who like hanging out, arguing through ideas, and turning unfinished thoughts into working prototypes. Krishna brings machine learning depth, Priyanka brings strong software engineering practice, and Shivam brings robotics and hardware instinct.
Software engineering
Priyanka
Product structure, clean systems, dependable interfaces, and the discipline that keeps prototypes from collapsing.
Robotics
Shivam
Motors, embedded control, mechanical tradeoffs, and the patience required to make physical things obey code.
Project 01
Tank: remote control for an Elegoo tracked robot.
The first experiment is deliberately concrete: take an Elegoo tank-style robot, make it controllable from a phone, then keep pushing. Local control, telemetry, internet relay, camera feedback, safety stop, autonomy experiments, and whatever breaks next.
01
Phone controls
Responsive driving controls that feel predictable on touch screens.
02
Internet relay
Remote operation without exposing the robot directly to the open web.
03
Safety loop
Command expiry, dead-man stop, speed limits, and visible robot state.
04
Learning path
A platform for perception, control, and small autonomy experiments.
The lab
Random ideas are allowed. Shipping small things is preferred.
Robotics
Small rovers, embedded controllers, sensors, telemetry, and human-in-the-loop control.
Machine learning
Perception, decision support, model-assisted tools, and practical evaluation.
Software
Interfaces, APIs, infrastructure, documentation, and apps that make experiments usable.
Odd ideas
The sketches that do not fit a category yet, but keep coming up during conversations.
Working style
Build the smallest honest version first.
- Make it real Prefer a working loop over a perfect plan.
- Keep it inspectable Document decisions, wiring, code paths, and failure modes.
- Respect the physical world Robots should fail by stopping, not by guessing.
- Stay playful The whole point is to learn together and enjoy the build.