Engineer & Builder

Hemanth
Vemulapalli

Data engineer, IoT builder, and founder of Sensoul — a privacy-first elderly care system powered by edge AI and sensor fusion.

Scroll to explore
5
Sensors fused
30K+
Samples logged
801
Features/sample

Building things that matter at the edge

I'm a data engineer and systems builder with a passion for turning hardware, data pipelines, and machine learning into real-world impact. I work at the intersection of embedded systems, edge computing, and AI.

Currently building Sensoul — a privacy-first elderly distress detection system that uses sensor fusion and edge ML to keep people safe without cameras or microphones.

📍 United States
🔧 Hardware to Cloud
From wiring breadboards and tuning I²C buses to building Parquet pipelines and ML-ready datasets — I own the full stack, end to end.
🧠 Privacy-First AI
Sensoul runs all inference on-device with no camera, no microphone, no cloud dependency. Edge ML that respects the people it serves.
📊 Data Engineering
Designing production-grade data pipelines with validation, compression, partitioning, and automated reporting — built to run for years unattended.

Sensoul — Privacy-First
Elderly Care System

A ceiling-mounted distress detection device that uses 5-sensor fusion to protect elderly individuals — with no camera, no microphone, and no cloud. All ML inference runs on-device on a Raspberry Pi 5.

When danger is detected, a weighted danger score triggers a 30-second cancel window before automatically alerting emergency services.

✅ All 5 sensors validated Raspberry Pi 5 Edge ML Sensor Fusion Privacy-First Python Parquet Pipeline
30K+
Samples
2 Hz
Sample rate
801
Features
MLX90640 Thermal Camera
32×24 pixel array · 110° FOV · Body heat detection
LD2410C mmWave Radar
UART · 256000 baud · Motion & stillness scoring
MH-Z19C CO₂ Sensor
UART2 · Breathing proxy · 374–732 ppm indoor range
MPU6050 IMU
I²C · Accelerometer + Gyroscope · Fall detection
VL53L1X Time-of-Flight
I²C · 0–4000 mm range · Floor distance & fall confirmation

Things I've built

01

Sensoul Data Pipeline

Production-grade 1-year continuous sensor logging pipeline. Parquet format, date partitioning, Great Expectations validation, and nightly Prefect orchestration with automated email reporting.

Python Parquet Prefect SQLite Pandas
02

Real-Time Sensor Dashboard

Live dashboard running on a 7" IPS display showing polar thermal radar visualization, multi-sensor status, and a weighted danger score bar — all updating at 2Hz from 5 simultaneous sensors.

Python Pygame Raspberry Pi 5 Threading
03

Edge ML Fall Detection

Training a multi-sensor fall detection classifier using tsfresh for automatic feature extraction and scikit-learn Random Forest — running entirely on-device with ONNX Runtime inference.

scikit-learn tsfresh ONNX NumPy
04

Remote Monitoring Stack

Zero-config global remote access for the Sensoul device using Tailscale VPN + Samba file sharing. Accessible from anywhere in the world including mobile data via iPhone Files app.

Tailscale Samba Linux NVMe

Skills & Tools

Hardware & Embedded
Raspberry Pi 5
I²C / UART / SPI
mmWave Radar
Sensor Fusion
Edge Computing
Data Engineering
Python
Pandas / PyArrow
Parquet / SQLite
Prefect / Airflow
Linux / systemd
ML & Dev Tools
scikit-learn
PyTorch
Git / GitHub
FastAPI
Docker
Want the full picture?
Download my resume for work history, education, and more.
↓ Download Resume

Let's build something together

Open to data engineering roles, IoT consulting, and conversations about Sensoul. Reach out anytime.