What we're building, what we learn on factory floors, and how we think about physical AI in India.
The LLM race is decided by GPU capital. The edge robotics race is decided by embodiment and deployments — a race India can win.
Read →An integration SI quotes weeks of engineer time for the same robot and machine. Our harness did it in 35 minutes. Here's how.
Read →Cloth folding at NTU, a Singapore apartment, and the pull of Indian factory owners — how Piqoid started.
Read →The editorial plan across four tracks — technical, customer-facing, company, and industry.
Mixture-of-tasks/experts: how task policies slot in as experts, why the vision backbone stays frozen, and how we keep policies verifiable.
Our RL setup end-to-end: Success Guided Sampling, why egocentric data can't teach hardware-specific motion, and what the tiny sim2real step actually does.
Plug in, and the brain reads torque, travel and specs over CAN. The hardware/software contract that kills PLC reverse-engineering.
Semi-deterministic completion — why most of a task should stay inspectable, and where we deliberately let RL take over.
Sensing the actual workpiece, thermal drift, fixtures that moved — what "unstructured 3D cutting & welding" means in practice.
Five iterations, 3+2 axes, reverse-SCARA — the design logic behind every deleted feature.
Thermal, dust, power quality, latency budgets, and why we refuse cloud dependency.
How we run a demo on the customer's own material — and why it's our entire sales process.
A day inside a factory running 120–130% of required labour — and what changed when the robot took the night shift.
Trust, references, payback fear, and the family dynamics behind the purchase — what we've learned watching buyers decide.
Correction by voice: the tacit knowledge that no dataset captures and how we harvest it on-site.
Why press-brake makers refer us — referral commissions, service economics, and a 15-year manual backlog.
What a robot actually costs to own — robot, integration, downtime, changes — and how to read a vendor quote.
The early prototypes, the first factory visit, and the decision to move to India full-time in June 2024.
Clusters beat campuses: the wedge logic of one task in one city.
Supply chains, tolerances, machine shops — what we've learned manufacturing here.
Our in-house design, sim and training stack — what we chose and why.
What we look for and why "ship to production" is the bar.
What we're learning about focus, cadence, and saying no — written as we go.
A recurring series: what shipped, what broke, what got cheaper.
Robot density India vs China vs global average — and why the missing piece isn't the arm.
Demographics, retirement, and why the next generation won't run manual factories.
Unit economics of engineer-hours vs compounding software, and the market gap it creates.
What the humanoid boom did for us (actuators) and what it doesn't solve (deployment).
Ownership of hardware + sim + deployments as the capital-light path — and why it generalises beyond India.
Policy, MSME incentives, and what a national robotics push should actually fund.
The best way to understand what we build is to watch it run on your own part.