Sandip Mondal·Software Engineer · QA Lead·West Bengal, India
I build things that hold up under pressure.
Software engineer and QA lead at Ralakde Automation UK. I write the code and then I try very hard to break it — Java and Kotlin on one side, Python and Selenium on the other.
One habit since my first line of C: build it, then find out how it breaks. Here is the short version.
I came to quality engineering the long way round — through Java, Android and a stubborn habit of taking things apart to find out why they broke.
Today I lead software quality at Ralakde Automation UK, where I own test strategy across the core application modules and drive the Python automation that keeps releases honest. I got there by building that framework myself: I joined as an SDET and wrote the end-to-end Selenium + Pytest suite from an empty repository.
The other half of my work is building. SkinGlance — an Android app that classifies skin lesions against a cloud-hosted CNN ensemble — is the project I am proudest of, because it had to be both: a model that is actually accurate and an app that does not fall over on a bad network.
The rest of the time I am in the hills. Hiking is where I do my clearest thinking, and the discipline turns out to be the same one good software asks for — pace yourself, trust the route, check your footing before you commit your weight.
“Simplicity is the soul of efficiency.”
The rule I hold myself to
The learning curve
2018
First lines of C
Where the curiosity started.
2022
Diploma
Computer Science & Technology · Murarai Government Polytechnic
2025
B.Tech
Computer Science & Engineering · Maulana Abul Kalam Azad University of Technology
2025
Teaching machines to look
Three CNNs, an attention-enhanced fuzzy ensemble, and a paper that survived peer review.
Dec 2025 →
Paid to break things
SDET, then QA Lead at Ralakde Automation UK.
Now
Still climbing
Deeper into everything I can't stop thinking about.
2018
First lines of C
Where the curiosity started.
2022
Diploma
Computer Science & Technology · Murarai Government Polytechnic
2025
B.Tech
Computer Science & Engineering · Maulana Abul Kalam Azad University of Technology
2025
Teaching machines to look
Three CNNs, an attention-enhanced fuzzy ensemble, and a paper that survived peer review.
Dec 2025 →
Paid to break things
SDET, then QA Lead at Ralakde Automation UK.
Now
Still climbing
Deeper into everything I can't stop thinking about.
The Ascent/1,240 m02 — 06
The ascent so far.
Ralakde Automation UK. I arrived to write the automation framework and stayed to lead the team that runs it.
Mar 2026 — Present● Current
Software Quality Assurance Lead
Ralakde Automation UK
Leading quality across the core application modules — strategy, people and the automation that backs both.
Manage team workflows and assign testing ownership across core application modules.
Drive test automation strategy in Python to keep UI updates robust release over release.
Streamline operations and raise software quality through cross-functional collaboration with development.
Python
Selenium
Pytest
Jira
UAT
Dec 2025 — Mar 2026
Software Development Engineer in Test
Ralakde Automation UK
Built the end-to-end UI automation framework the team still runs on, starting from an empty repository.
Built an end-to-end UI automation framework using Python, Selenium and Pytest on the Page Object Model.
Ran user acceptance testing for critical application components, verifying functionality and system compliance.
Optimised testing pipelines and troubleshot complex configuration issues across environments.
Python
Selenium
Pytest
Page Object Model
REST APIs
The Traverse/2,680 m03 — 06
Things I have built and broken.
Everything here is honestly described. The first two are the ones I would ask you to look at — and the research behind them is published.
An Android app that detects potential skin cancer from a lesion photo. Capture or pick an image, crop it live in the camera preview, and get a benign/malignant call with a confidence score from a cloud-hosted model.
The interesting problem was not the model — it was everything around it. Lesion photos need precise framing, so the app crops in real time against a custom DrawView overlay before a byte ever leaves the phone. Retrofit handles the round trip, and every failure mode surfaces as a readable message instead of a crash.
The brain behind SkinGlance: three CNNs — DenseNet169, InceptionV3 and MobileNetV2 — trained on the ISIC dataset and combined through a fuzzy ensemble, served as a REST API.
No single architecture was reliably right, so none of them decides alone. A fuzzy ensemble weighs all three outputs into one prediction, and FastAPI serves it over REST so the Android client stays thin. Training runs locally or straight from a Kaggle notebook.
I am open to software engineering and SDET roles, and always happy to talk about test automation, Android, or a route worth walking. Write to me — I reply to everything that is not a recruiter template.