I’m Sreekar, a Class 11 student in Hyderabad. I’m drawn to electronics, embedded hardware, and writing software that solves real, practical problems.
This notebook is my personal engineering journal. Rather than presenting only tidy final projects, it documents what actually happens at the workbench: the noisy breadboard connections, the calibration constants that didn't make sense at first, the code that failed before it worked, and the questions I brought to my mentors.
The Four Projects
Over nine months—from June 2025 to February 2026—I worked across four long-running projects spanning AC power measurement, power electronics, edge machine learning, and community e-commerce architecture:
1. Power Quality Recorder
An AC power-monitoring system and data logger built from scratch. Starting from basic principles of AC waveforms, this project evolved into a multi-sensor setup using an ESP32 microcontroller, an ADS1115 16-bit external ADC over I2C, a ZMPT101B voltage transformer, and an ACS712 Hall-effect current sensor. The logs follow the journey from noisy breadboard lines to a soldered protoboard, implementing discrete Root-Mean-Square (RMS) sampling loops, calculating true active power (Watts) versus apparent power (VA), determining power factor (φ), establishing noise thresholds below 0.05 A, and detecting AC zero-crossings for frequency tracking.
2. DC-DC Buck Converter Comparison
A hands-on comparative analysis of two popular step-down switching regulators: the asynchronous XL4015 (with an external Schottky diode) and the synchronous MP2307 (with integrated MOSFETs). Moving past optimistic datasheet claims like “up to 96% efficiency,” I measured how switching frequency (180 kHz vs 340 kHz) impacts inductor ripple, calculated forward conduction diode losses, and observed thermal throttling under sustained current draw.
3. TinyML Audio Keyword Spotter
Bringing artificial intelligence to low-power edge microcontrollers. Using an I2S digital MEMS microphone connected to an MCU, I developed a lightweight voice-command detection pipeline targeting two keywords (“on” and “off”). The notes cover real-time audio sampling, on-device Mel-Frequency Cepstral Coefficients (MFCC) feature extraction, confusion matrix evaluation, and measuring the real energy cost per inference (46.3 mJ) to understand sleep duty cycles and battery life.
4. Kindred Hands Platform
A collaborative e-commerce platform and digital directory created to empower traditional handloom weavers in Andhra Pradesh and Telangana. This track focuses on technical and economic pragmatism: comparing monolithic multi-vendor stacks (WordPress, WooCommerce, and Dokan) against custom, lean open-source solutions. The goal was to eliminate high recurring plugin renewal fees, simplify mobile onboarding for non-technical artisans, and map out realistic catalog workflows and order fulfillment.
How I Work & Learn
- First principles over shortcuts: If a formula includes a mysterious multiplier or an offset like 1.65 V, I trace where it comes from before letting the code run.
- Honest data: Physical hardware comes with noise, component tolerances, and drift. The logs record what the instruments actually measured, not idealized numbers.
- Mentorship & iteration: When hitting a wall, I focus on asking clear, bounded questions with my circuit diagrams and code open, learning how to debug methodically.
About the Dates & Sequence
The 82 entries in this notebook are presented in the chronological order of the build milestones, experiments, and architectural pivots as they unfolded between June 2025 and February 2026. This digital edition was compiled in October 2026 to bring the four related project logs into one unified reference notebook.