The Future of Robotics: Building Machines That Make More Possible
Ajitsingh Jagtap — Founding Mechanical Engineer of a Stealth Robotics startup
Ajitsingh Jagtap is the founding mechanical engineer of a stealth robotics startup building autonomous robots that automate manual manufacturing tasks. His central lesson in robotics engineering is that reliability, not generality, is the hard part: a robot that does one useful thing ten thousand times without intervention is worth far more than one that does a hundred things once in a demo video.
Ajit's company is building autonomous robots that automate manual manufacturing tasks. The thesis centres on throughput: raising production throughput across every vertical by orders of magnitude, to the point where artificial scarcity can no longer be engineered. The team calls the end state an Exafab, a modern-day matter compiler, or a machine that can make anything.
Rather than starting from nanotechnology, the company starts from advanced robotics, automating manual labor in critical industries and increasing the throughput of matter. Its "the factory is the product" mindset, applied across all goods and services and scaled aggressively, is how the team believes it can approach the ideal matter compiler asymptotically.
Ajit works on the anthropomorphic hand team, designing the palm and finger mechanisms for dexterous manipulation in data centers. In his view, drawing the mechanism is not the hard part. A hand has to work on parts that don't cooperate, in tolerances that drift, with forces that can't be fully predicted.
His answer is discipline: get a rough, instrumented build in front of real failure modes early, before committing to an architecture that only looks good on paper.
Dexterity makes this question sharp, Ajit explains, because the mechanism cannot be separated from the control. His team uses a compliant palm, so some of the intelligence lives in the mechanics rather than the code. The hand absorbs misalignment physically instead of requiring perfect perception.
The general principle is to decide what the structure should solve and what the software should solve, and never ask either to compensate for a bad decision in the other.
According to Ajit, no startup gets everything right at the beginning, and designing as though it will is its own failure mode. The real question is what to make cheap to be wrong about.
A few decisions are expensive to change later: the kinematics, how subsystems mount and route, and how serviceable the machine is when it fails at 2 a.m. These deserve careful attention early.
Everything else is deliberately left loose, because the first build's job is to reveal what the team didn't know. The common mistake Ajit sees is polishing surfaces before the architecture is settled, which produces something that looks finished and teaches nothing.
Ajit points to manufacturing and critical infrastructure, where the labor gap is real and the economics already work. The biggest misunderstanding, he says, is that generality is the hard part. The hard part is reliability.
Comments
No comments yet — be the first to share your thoughts.