Poovarasan V.

• Developing a non-invasive optical bilirubin detection device focused on improving neonatal healthcare.
• Designing multi-wavelength LED–photodiode sensor circuits for accurate optical measurement and signal analysis.
• Collaborating on hardware design, prototyping, and system testing to validate device performance.

• Built hands-on embedded system projects and strengthened skills in debugging, testing, and optimizing embedded code.
• Worked on hardware–software integration using Microchip development modules and industry-grade tools.
• Gained practical exposure to microcontroller architectures and real-world embedded workflows.

The project involved designing and fabricating a custom PCB-based optical probe incorporating LEDs at 460 nm, 540 nm, 660 nm, and 950 nm to enable wavelength-specific tissue characterization. The team implemented multiple source–detector separations ranging from 3.40 mm to 9.70 mm to study depth-dependent light–tissue interactions. A complete optical measurement system was developed using a BPW34 photodiode, LM358 transimpedance amplifier, and Arduino microcontroller for controlled LED activation and signal acquisition. The system was evaluated using a skin-mimicking phantom, demonstrating consistent and predictable reflectance behavior across varying depths, confirming the probe’s effectiveness for non-invasive tissue analysis.

A hardware development platform built around the Texas Instruments MSPM0G3507 ARM Cortex-M0+ microcontroller, designed with professional PCB layout practices using KiCad. The board features USB-C power and data, integrated 3-axis MEMS accelerometer (LIS2DH12) for sensing applications, USB-to-UART bridge (CH340E) for communication, and a Tag-Connect SWD footprint for debugging. Power management is handled through a precise 3.3 V LDO regulator, and the schematic and PCB layout prioritize clean signal routing and usability. This project demonstrates complete end-to-end custom board design—from schematic to PCB—enabling rapid prototyping and embedded systems experimentation.

As a team, we developed a low-cost predictive maintenance system that uses an ESP32 microcontroller to create a real-time digital twin of industrial machinery. The setup continuously monitors vibration, temperature, and current using the MPU6050, thermistor, and ACS712 sensors, and runs a TinyML regression model directly on the ESP32 to predict healthy operating behavior. By comparing real-time sensor readings with digital-twin predictions, the system accurately identifies anomalies such as imbalance, overheating, or electrical overload without requiring cloud computing. The project demonstrates efficient edge-based fault detection with ~96% accuracy, low latency (~180 ms), and minimal power consumption, offering an affordable Industry 4.0 maintenance solution for small-scale industries.

This project focuses on designing and implementing a Class-D switching amplifier, an audio amplification system that uses high-frequency PWM (pulse-width modulation) to achieve high efficiency and low power loss. The amplifier operates by converting the input audio signal into a PWM waveform, driving switching MOSFETs that operate fully ON or OFF, significantly reducing heat dissipation compared to linear amplifiers. The amplified PWM signal is then passed through an LC low-pass filter, which reconstructs the original audio waveform at a higher power level, enabling clear and efficient sound output. This design highlights the principles of switching amplification, efficient power conversion, and practical audio system implementation.

This project involved building an automated real estate rental web scraper that collects key property details such as pricing, location, and amenities using BeautifulSoup and Selenium. The scraper navigates through listing pages, extracts structured data, and exports the results directly into Google Sheets for easy access and analysis. This system significantly improves the efficiency and accuracy of rental data collection, making it useful for market research, trend analysis, and personal decision-making.

This project involved building an automated real estate rental web scraper that collects key property details such as pricing, location, and amenities using BeautifulSoup and Selenium. The scraper navigates through listing pages, extracts structured data, and exports the results directly into Google Sheets for easy access and analysis. This system significantly improves the efficiency and accuracy of rental data collection, making it useful for market research, trend analysis, and personal decision-making.

EMAIL
poov4rasan@gmail.com