p.
✳ Computer Science Student PORTFOLIO / 2026

Pavan Kumar Goud Amikula

Curious by nature. Driven to solve.

I'm a Computer Science master's student at BTH. I turn complex questions into experiments, compare the evidence, and build usable applications. I'm eager to learn, collaborate, and work through challenging problems.

Explore my work GitHub
01 — EXPLORECode. Curiosity. Possibility.
INTERACTIVE FIELD / 01
EXPLORINGIntelligence

Drag to rotate · Choose a field to transform

Machine learningComputer visionBackend & cloudFROM EXPERIMENTS TO APPLICATIONS
01 / SELECTED WORK

Ideas, built
into reality.

A selection across intelligent systems,
explainability, and application development.

V4FINBENCH / MODEL EVALUATION
V4FinBench model comparison chart from the project results
RESEARCH IN PRACTICEFIG. 01
Machine learning01

V4FinBench

Financial distress prediction under extreme class imbalance, with evaluation across companies and over time.

Problem, approach & my contribution
The problem
Rare distress events and time-dependent data can make a strong-looking model unreliable.
The approach
Compare boosted models using company-grouped validation, temporal testing, feature-timing audits, and SHAP.
My contribution
A portfolio benchmark exploring careful evaluation, leakage risks, and explainability.
LightGBMXGBoostCatBoostSHAP
Explore repository
Machine learning02

Heavy Vehicle Predictive Maintenance

Exploring rare vehicle-failure classification through synthetic telemetry, model comparisons, and local explanations.

Problem, approach & my contribution
The problem
Missed failures can be more costly than false alarms, while failure examples are scarce.
The approach
Aggregate sensor features, handle imbalance with SMOTE, compare classifiers and ensembles, evaluate error costs, and connect predictions to a Flask interface.
My contribution
Coauthored the academic report with Naga Sreerama Pradyumna Tata. The report documents the shared pipeline; it does not assign individual implementation tasks.
PROJECT REPORTSPDFPredictive maintenance report
scikit-learnSMOTELIMEFlask
Explore repository
Computer vision03

Brain Tumor Classification

Comparing CNNs and ensemble explanations to understand MRI classification beyond overall accuracy.

Problem, approach & my contribution
The problem
High aggregate accuracy can hide class-specific errors and the trade-off between efficiency and reliability.
The approach
Compare CNN architectures with per-class metrics and confusion matrices, then investigate ensemble learning and Grad-CAM explanations.
My contribution
Implemented and trained InceptionV3 and MobileNetV3 in the CNN comparison. Jointly contributed preprocessing, evaluation, confusion matrix analysis, and writing. Also a named major-project team member and ensemble-study coauthor.
PROJECT REPORTSPDFCNN comparison · my implementation rolePDFEnsemble & explainable AI studyPDFMajor-project team report
TensorFlowKerasInceptionV3MobileNetV3Grad-CAM
Explore repository
Cloud & backend04

IncidentLab

Exploring cloud incident detection using multimodal telemetry and a model-serving API.

Problem, approach & my contribution
The problem
Signals from different parts of a cloud system can be difficult to connect during an incident.
The approach
Explore GRU and Isolation Forest models alongside a FastAPI serving layer and containerized infrastructure.
My contribution
A portfolio project connecting telemetry analysis, model serving, and cloud tooling.
PyTorchFastAPIDockerKubernetes
Explore repository
Cloud & backend05

CampusCart

An equipment-reservation application built around clear APIs and containerized microservices.

Problem, approach & my contribution
The problem
Equipment reservations need clear availability and consistent application workflows.
The approach
Organize the application into services with FastAPI, PostgreSQL, Docker, and Kubernetes.
My contribution
A portfolio application exploring service boundaries, data, and deployment.
FastAPIPostgreSQLDockerKubernetes
Explore repository
02 / THE PERSON BEHIND THE CODE

Always a student.
Always building.

I like understanding why a system works, where it fails, and how to make it useful. My projects connect machine learning, computer vision, explainability, and backend development. In my CNN architecture comparison, I implemented and trained InceptionV3 and MobileNetV3, and worked with my coauthor on preprocessing, evaluation, confusion matrix analysis, and writing. I bring that same curiosity to new problems: ask good questions, test alternatives, learn from the results, and keep improving.

EDUCATIONM.Sc. Computer Science · Blekinge Institute of Technology · Sweden
MY TOOLBOX

Tools I work with.

PythonSQLNumPypandasscikit-learnTensorFlowKerasPyTorchLightGBMXGBoostSHAPLIMEGrad-CAMFastAPIFlaskPostgreSQLDockerKubernetesGit

Applied across the projects above, from experimentation to deployment.

03 / PROFESSIONAL EXPERIENCE

Learning beyond
the classroom.

May 13, 2025 – August 12, 2025

Intern

Nukkad Shops Technologies (India) Private Limited

Completed an internship, bringing my learning into a professional team environment. The completion certificate confirms my role and dates.

PDFView completion certificate
04 / LET'S CONNECT

Next great idea?
Let's talk.

Explore my code, follow my learning journey,
or connect about a project.

Find me on GitHub