Projects shipped.
Problems solved.

EaseEmbro Production Suite

A complete AI + software + hardware production management system for embroidery factories.

EaseEmbro began as a problem we lived. Operating an embroidery factory with Dahao machines, we saw time lost to manual production counting and payroll calculation. We built the complete system: IoT sensor nodes mounted on each machine count output in real-time, push to cloud data pipelines, and feed a live dashboard for production, job orders, workers, and payroll calculations including advances, deductions, and shift adjustments.

Key technical decisions

• Sensor type: Proximity + IR for non-contact machine counting
• Communication: WiFi with local fallback buffer for connectivity drops
• Database: Real-time sync with offline-first architecture
• Payroll engine: Rule-based calculation with piece-rate, advance, and deduction handling
• Deployment: Cloud-hosted and accessible via browser on any device

SmartAttend Biometric Terminal

A ruggedized fingerprint attendance device designed and built for factory environments.

SmartAttend was designed for industrial conditions, not office assumptions. The enclosure is dust-resistant and suited for continuous operation in textile factories. Fingerprint events sync to cloud in seconds and integrate directly with EaseEmbro attendance and payroll workflows, eliminating manual attendance sheets.

Key technical decisions

• MCU: ESP32 for WiFi-native low-power continuous operation
• Sensor: Optical fingerprint module with 1:N matching up to 200 workers
• Enclosure: Custom 3D-designed housing with factory-grade material
• Connectivity: WiFi with local queue for network-loss scenarios
• Power: 5V USB-C input with surge protection circuit

IndustrialEdge

IoT Gateway Reference Design

An edge gateway that bridges legacy industrial machines to cloud-ready data pipelines.

IndustrialEdge was designed to connect RS-485 and Modbus RTU machine data to modern software systems. The gateway reads and buffers data locally, then publishes to MQTT with retry logic for unstable networks. Deliverables included schematic, PCB, firmware, and deployment architecture as a practical reference design.

Key technical decisions

• Protocol: Modbus RTU master in custom firmware
• Physical layer: RS-485 transceiver with industrial isolation
• MCU: STM32F4 for deterministic real-time polling and buffering
• Cloud transport: MQTT over TLS with retry and retained states

48V / 15A E-Bike Battery Management System

A production-ready 13S Li-ion BMS design for 48V e-bike battery packs.

This project focused on building a reliable 48V / 15A BMS for 13S Li-ion packs with production intent. The architecture uses BQ76930-based sensing and dual MOSFET protection, with a 2oz copper PCB and Kelvin current measurement for stable high-current operation. Along with the circuit design, we defined a repeatable manufacturing and QA flow for deployment.

Key technical decisions

• AFE: BQ76930 with native 13S support and passive balancing
• Protection: Dual IRFB4110 MOSFET topology for charge/discharge safety
• Current sensing: Low-ohm shunt with 4-wire Kelvin routing
• Thermal monitoring: Pack-side NTC feedback for cutoff logic
• PCB: 2oz copper power stage with high-current trace and thermal design

Automated Diamond Polishing System

A closed-loop STM32 system for precision facet polishing with ADC-based quality detection.

Surat's polishing workflow still depends heavily on artisan feel and manual judgement. We built a closed-loop embedded system where a geared stepper positions the polishing head and an ADC-based sensor circuit detects facet quality in real time. When the target threshold is reached, firmware executes raise-and-advance automatically. The goal is repeatable facet quality from programmable logic instead of operator variability.

Key technical decisions

• MCU: STM32 for deterministic real-time step/direction pulse control
• Motion: Planetary gearbox + microstepping driver for sub-degree resolution
• Feedback: ADC channel sampling analog surface finish signal
• Threshold logic: Configurable per facet geometry and quality target
• State machine: Position → Contact → Sample → Raise → Next facet

Embroidery Machine IoT Production Monitor

A factory-floor IoT monitoring system with local MQTT, live OEE dashboards, and zero cloud dependency.

This deployment was built for factories where internet reliability cannot be assumed. Each machine uses an ESP32 node with four sensors for RPM, current, thread-break, and temperature, then publishes telemetry every second over local MQTT. An Android tablet runs the broker and dashboard stack, stores data in SQLite, and provides live machine status, shift summaries, and OEE metrics without cloud dependency.

Key technical decisions

• MCU: ESP32-WROOM for WiFi-native multi-sensor acquisition on each machine
• Power: Isolated AC-DC module for direct 220V machine-side installation
• Sensors: Hall + CT + optical + NTC combined into a 1s telemetry packet
• Broker stack: Mosquitto embedded on Android tablet with local persistence
• Resilience: Auto reconnect, local queueing, and offline-first operation