DEV Community: remontsuri

EV-QA-Framework: Open-Source ML-Powered Quality Analysis for EV Battery Systems

remontsuri — Fri, 29 May 2026 08:54:10 +0000

EV-QA-Framework: Open-Source ML-Powered Quality Analysis for EV Battery Systems

EV-QA-Framework is an open-source machine learning framework for analyzing electric vehicle battery data. Whether you work with BMS telemetry, Tesla Model S/X logs, or just want to experiment with ML in the automotive domain — this tool covers the full pipeline from data ingestion to visualization.

What's Inside

1. Telemetry Validation

Raw CAN bus data is noisy. The validation module checks:

Temperature ranges per cell
Voltage-SOC correlation
Timestamp continuity
Physical consistency of readings

from ev_qa_framework import QAFramework

qa = QAFramework()
report = qa.validate_telemetry("battery_data.csv")
print(report.summary())

2. Anomaly Detection (Isolation Forest)

Automatically finds outliers in battery telemetry — early indicators of cell degradation or defects:

from ev_qa_framework.analysis import BatteryAnomalyDetector

detector = BatteryAnomalyDetector(contamination=0.05)
anomalies = detector.fit_predict(telemetry_data)
detector.visualize()

The visualization highlights points where battery behavior deviates from normal patterns — critical for predictive diagnostics.

3. SOH Prediction (LSTM)

Predicts State of Health (remaining capacity) from historical charge/discharge cycles:

from ev_qa_framework.soh_predictor import SOHPredictor

predictor = SOHPredictor(seq_length=50)
predictor.train(charging_cycles)
predicted_soh = predictor.predict(current_data)
print(f"Predicted SOH: {predicted_soh:.1f}%")

The model supports export for embedded deployment.

4. CAN Bus Emulator

Test your pipeline without hardware — generates realistic battery telemetry:

from ev_qa_framework.can_bus import CANEmulator
from ev_qa_framework.config import TESLA_MODEL_S_CONFIG

emulator = CANEmulator(config=TESLA_MODEL_S_CONFIG)
for frame in emulator.stream(frequency=10):
    dashboard.update(frame)

Supports custom load profiles via settings.yaml.

Web Dashboard

Built with FastAPI — real-time telemetry visualization out of the box:

docker compose up -d
# or directly:
pip install ev-qa-framework
python quickstart.py

The dashboard shows temperature, voltage, SOC graphs, and detected anomalies.

Quick Start

pip install ev-qa-framework
python -c "
from ev_qa_framework import QAFramework
qa = QAFramework()
qa.quickstart()
"

Or with Docker:

docker pull ghcr.io/remontsuri/ev-qa-framework:latest
docker compose up

Links

MIT licensed. Contributions, issues, and discussions are welcome.

Building an ML-Powered Battery Testing Framework for Electric Vehicles

remontsuri — Mon, 19 Jan 2026 22:47:32 +0000

🔋 Building an ML-Powered Battery Testing Framework for Electric Vehicles

Battery failures cost the EV industry $5 billion annually in warranty claims, recalls, and safety incidents. A single thermal runaway event can destroy an entire vehicle and endanger lives.

The problem? Millions of telemetry data points (voltage, temperature, State of Charge) go unmonitored until it's too late. Manual QA can't scale.

I built EV-QA-Framework — an open-source Python testing framework with AI-powered anomaly detection for Battery Management Systems (BMS).

GitHub: https://github.com/remontsuri/EV-QA-Framework

🛠 Why I built this

Mainstream testing tools are great for web apps, but they lack native support for hardware-specific edge cases like:

Voltage Drop Anomaly: Detecting subtle dips that indicate cell degradation.
Rapid Discharge Spikes: Identifying patterns before they lead to short circuits.
State-of-Charge (SoC) Drift: Ensuring the reported battery percentage is actually accurate.

🚀 Key Features

64+ Automated Tests: Covers voltage, current, temperature, and charging cycles.
ML Anomaly Detection: Uses scikit-learn's IsolationForest to spot outliers in telemetry data that rule-based systems miss.
High Performance: Built with Python 3.12, Pydantic (v2) for ultra-fast validation, and Pytest.
Interactive Reports: Generates Jupyter Notebook visualizations for engineers to "see" the failure.

📊 Quick Demo

Here is how it looks in action (Full demo in the repo):

from ev_qa_framework.ml.anomaly_detector import BatteryAnomalyDetector

# Load telemetry data
detector = BatteryAnomalyDetector()
anomalies = detector.detect_anomalies(telemetry_df)

# Visualize findings
detector.plot_results(telemetry_df, anomalies)

🌍 Open Source & Community

This project is open-source because EV safety should be accessible. Whether you are a QA engineer at Tesla or a hobbyist building a DIY e-bike, I hope this helps you build safer products.

I'm currently looking for new opportunities in the EV/Green-Tech industry as a QA or ML Engineer. If your team is hiring, let's connect!

Check it out here: https://github.com/remontsuri/EV-QA-Framework

python #machinelearning #testing #ev #opensource

Building PsychoAnalyze AI: An AI-Powered Interview Analysis Tool for Therapists

remontsuri — Wed, 10 Dec 2025 21:10:54 +0000

I built PsychoAnalyze AI, an open-source tool that combines my background in psychology with full-stack development to help therapists and researchers streamline their interview analysis workflow.

Problem

Therapists and researchers spend significant time analyzing psychological interviews manually. The process is time-consuming and requires expertise in identifying patterns related to:

Defense mechanisms
Attachment styles
Emotional responses

Solution

PsychoAnalyze AI leverages Google Gemini API to:

Detect psychological patterns automatically
Identify attachment styles from interview transcripts
Recognize emotions and emotional patterns
Provide therapy-oriented recommendations
Support 6 languages
Export analysis to PDF/Excel

Tech Stack

React + TypeScript
Vite
Tailwind CSS
Google Gemini API

Try It Out

Live demo: https://psychoanalyze-ai.vercel.app
GitHub: https://github.com/remontsuri/psychoanalyze-ai

Looking for feedback from the dev community! Any suggestions for improving the tool or adding new features?