Research notes, engineering stories, and thinking from the team. We write when we have something real to say.
Bipedal locomotion is deceptively hard. Getting a robot to stand is one problem — getting it to recover from an unplanned shift in weight distribution is a completely different one. We document the sensor feedback loops, the failed early prototypes, and the moment coordinated movement finally clicked in our lab.
Most industrial monitoring systems tell you something broke. PHMS is designed to tell you what's about to break — and why. Here's the architecture decision that changed everything: treating sensor drift as signal, not noise.
Industrial drive systems are often installed and forgotten. But a well-tuned drive with the right ramp profiles, harmonic filtering, and closed-loop feedback can cut energy consumption by 30–40% on the same load. We share how we configure them on industrial deployments.
We're a small team building hard things. We don't hire for pedigree — we hire for curiosity, first-principles thinking, and the ability to move from whiteboard to working system without needing a playbook. What that looks like in practice.
Running inference on embedded hardware is about more than FLOPS and memory bandwidth. Thermal limits, interrupt latency, and power rail noise all affect model reliability in ways that no benchmark captures. Lessons from a year of shipping real hardware.
Building robotics and industrial AI in India in 2025 means navigating gaps in supply chains, limited local hardware ecosystems, and a talent market that skews toward software. Here's how those constraints pushed us to build leaner and smarter.
We write when we have something real to say. No spam — just research notes, engineering stories, and honest updates from the lab.
No account needed. Unsubscribe anytime.