EmbedCrest Technology Private Limited

OBJECTIVE:

 To build a visually rich, standalone IoT environmental monitoring station capable of real-time temperature and humidity display with professional-grade embedded graphics and optimized ESP32 firmware.

ACTION:

Developed ESP32 C++ firmware integrating the DHT11 sensor and SSD1351 OLED/TFT display. Implemented SPIFFS for filesystem support and used a JPEG decoding engine to render custom graphics (temperature and humidity icons) dynamically.

 Key technical contributions:

• High-resolution UI rendering on embedded hardware 
• Sensor acquisition with noise filtering and stable sampling logic 
• Efficient SPI display refresh and memory
• optimized asset handling
• Modular firmware architecture for easy feature expansion 

RESULT:

Delivered a robust, visually appealing IoT station with accurate real-time data processing and smooth embedded graphics. The solution demonstrates strong capabilities in ESP32 firmware engineering, sensor integration, UI/UX on microcontrollers, and low-power IoT device development.

OBJECTIVE:

 To design a reliable, microcontroller-based electronic voting machine using ATmega32 capable of capturing user votes, ensuring secure confirmation, providing real-time LCD feedback, and managing system states with clear LED indicators.

 ACTION:

 Developed complete embedded firmware and hardware integration around the ATmega32 MCU. Key engineering work included:

• Implemented button-driven voting inputs using PB0–PB4
• Added Responsible Mode (PA0) and Reset Function (PA1) for secure vote management
• Integrated an LCD (LM016L) to display candidate vote counts continuously
• Controlled multiple status LEDs:
• Candidate selection LEDs (PC0–PC3) 
• Responsible-mode LEDs (PC4, PC5)
• System status/ready LED (PC6) 
• Built a state-machine architecture to manage voting, confirmation, resetting, and user prompts 
• Implemented logic to ensure only one candidate is active at a time and prevent invalid inputs
• Designed structured LCD output format:
• Added protection logic: “please vote” prompt if enter is pressed without selecting a candidate

 RESULT:

 Delivered a fully functional, user-friendly electronic voting system demonstrating accurate vote counting, secure confirmation workflow, real-time visual feedback, and robust embedded firmware design. Suitable for education, labs, and demonstrative e-governance simulations.

OBJECTIVE:

To build a complete smart home automation system using dual ATmega32 microcontrollers with centralized control over lights, fans, AC, TV, door lock, and multi-room monitoring. The aim was to deliver a low-cost, reliable, and expandable home automation platform with secure user access.

ACTION:

Designed and programmed a full master–slave architecture using SPI communication between two ATmega32 MCUs.

Key engineering tasks included:

• Implemented Guest/Admin access modes with keypad-based input and LCD UI
• Integrated LM35 temperature sensor for real-time AC control
• Added EEPROM persistence for restoring device states after reset
• Developed control logic for lights, fans, TV, AC, and servo-based door lock
• Used L293D motor drivers for motor/fan operations
• Built room modules (Room1–Room4) with individual fan/light controllers
• Designed a clean user workflow for authentication, device control, and monitoring

RESULT:

Delivered a complete smart home controller capable of managing multiple rooms, regulating temperature automatically, and securely operating all household appliances from a single interface. Demonstrated strong expertise in embedded system design, master–slave communication, hardware integration, and user-centric HMI development.

The Smart Healthcare Prediction System is an intelligent medical assistant that uses machine learning to predict the chances of having a disease based on a patient’s health data such as age, blood pressure, glucose level, BMI, and symptoms. The system analyzes medical records and provides quick health risk predictions that help doctors and patients make early decisions for treatment. It reduces manual effort, improves accuracy in diagnosis, and supports preventive healthcare. The system is deployed as a simple web application where users can enter their details and instantly get a result indicating whether they are at high or low risk of a particular disease.

Fast and accurate disease risk prediction User-friendly web interface Doctors and patients can use it easily Helps detect disease at an early stage Supports better healthcare decision-making.

Python, Machine Learning Natural Language Processing Scikit-Learn, Pandas, NumPy Flask For Deployment

The Smart Healthcare Prediction System successfully predicts the health risk of a patient with high accuracy based on their medical data. After testing multiple machine learning models, the Random Forest Classifier performed the best with an accuracy of around 82-86% . High Risk → “Immediate medical consultation recommended.” Low Risk → “Healthy condition. Continue regular checkups.” The predictions were validated using real patient datasets, and performance was measured using metrics like: Accuracy the system proved to be fast, reliable, and useful for preventive healthcare.

The Vehicle Insurance Prediction System is a machine learning–based solution designed to assist insurance companies in predicting whether a customer is likely to purchase vehicle insurance or raise a claim. Insurance companies handle huge amounts of customer data, and making decisions manually can be time-consuming and inaccurate. This project uses data analytics and predictive modeling to improve decision-making in the insurance sector.

OBJECTIVE :-

To build a smart model that predicts: Whether a user will buy vehicle insurance Helps insurance companies target potential customers Improves business strategy and reduces marketing cost Data Cleaning, Pre-processing & Feature Engineering Exploratory Data Analysis (EDA) for business insights Model Training using ML classification algorithms Hyperparameter tuning for maximum accuracy Performance evaluation using metrics (Accuracy, Precision, Recall, F1-Score, ROC-AUC) Flask web app deployment

RESULT :-

Built a high-accuracy prediction system (> 85% accuracy depending on model) Helps insurance companies target customers more effectively Reduces operational and marketing costs

OBJECTIVE:

To build a dual-role BLE communication system on ESP32 capable of scanning, connecting, serving notifications, and displaying real-time data through a responsive TFT UI. The goal was a robust, user-interactive IoT connector supporting both BLE Client and BLE Server roles.

ACTION:

• Developed C++ firmware on ESP32 implementing:
• BLE Server (Notification TX) for sending characteristic updates
• BLE Client (Scan / Connect / Read / Write) for interacting with external BLE peripherals Integrated a TFT UI with button-driven navigation to:
• Scan and list up to 10 BLE devices
• Select a target peripheral
• Display connection status and live characteristic values Firmware included optimized BLE event handling, non-blocking UI refresh logic, and a clean modular architecture for easy integration into larger IoT systems.

RESULT:

Successfully delivered a fully functional, real-time BLE dual-role connector with smooth UI interaction and reliable characteristic exchange. The system enables rapid prototyping and deployment of BLE-based IoT products, user interfaces, and connectivity testing tools.

OBJECTIVE: To benchmark true dual-core concurrency across different microcontroller platforms (ESP32-S3 and Portenta H7) under identical heavy computational workloads. The aim was to validate parallel execution efficiency, measure core utilization, and compare real-world multicore performance for advanced embedded applications. ACTION: I engineered two high-load computational tasks (Prime search and Pi calculation) and executed them using FreeRTOS with strict core-pinning to enforce deterministic concurrency. A cross-platform benchmarking framework was developed to: -Run identical workloads on ESP32-S3 and Portenta H7 -Measure execution time per core -Display real-time load and progress on TFT displays -Capture performance differences between architectures This included optimized RTOS scheduling, deterministic task assignment, and a custom visualization layer. RESULT: The system successfully demonstrated 100% dual-core utilization, verified true parallel execution, and revealed significant performance variations between MCUs. The benchmarking solution provides a reliable method to evaluate multicore behavior for robotics, edge AI, industrial control, and compute-heavy embedded systems requiring deterministic concurrency.