Featured Work

Production Systems & Engineering Projects

Enterprise ERP automation deployed on Linux VMs with Docker & GitHub Actions CI/CD, observability stacks, LLM/RAG assistants, and data warehouse pipelines.

Visionflow AI (KTH Automation)
SRE & DevOpsMar 2026 — Sep 2026
Production Linux VM • Automated CI/CD • NPME Silver Award
Silver Award — Best Managed Project in Process Automation (NPME Awards 2026)

Visionflow AI (KTH Automation)

Production ERP & Automated Data Reconciliation System (with Sri Lanka Telecom)

Developed Visionflow AI, an advanced automated ERP and data reconciliation platform for KTH Tyre House in collaboration with Sri Lanka Telecom (SLT). Streamlines complex Sales Summary Reporting and Supplier Invoice Matching workflows, deployed as a containerized production workload on Linux with automated CI/CD and HTTPS.

  • Containerized multi-service architecture deployed on a production Linux VM via Docker and Docker Compose.
  • Automated build and deployment workflow engineered with GitHub Actions CI/CD pipelines.
  • Secured production endpoints with Let's Encrypt SSL/TLS certificates managed via Certbot.
  • Maintained high reliability through container log observability and PostgreSQL schema/query optimization.
DockerDocker ComposeGitHub Actions CI/CDLinux VMPostgreSQLLet's EncryptCertbotPythonFastAPI
Linux & Container Observability Stack
SRE & DevOps2026
Real-Time Metrics • Host & Container Telemetry • Custom Dashboards

Linux & Container Observability Stack

Real-Time Infrastructure Telemetry with Prometheus, Node Exporter & Grafana (Editable Showcase)

An end-to-end infrastructure monitoring and alerting blueprint designed for Linux virtual machines and Dockerized workloads. Collects host-level hardware and OS metrics via Node Exporter, scrapes time-series telemetry with Prometheus, and visualizes golden signals (CPU, memory saturation, disk I/O, network throughput) in Grafana.

  • Declarative Docker Compose deployment spinning up Prometheus, Node Exporter, and Grafana with persistent volumes.
  • Pre-configured scrape targets and health checks for monitoring Linux host saturation and container uptime.
  • Custom Grafana dashboards tracking SRE golden signals: latency, traffic, errors, and resource saturation.
PrometheusGrafanaNode ExporterDocker ComposeLinux CLIObservabilitySRE
Source Code
AI-Powered E-Commerce Chatbot — RAG & Text-to-SQL
AI & BackendDec 2025 — Jan 2026
643 Product Catalogue • Llama 3.3 via Groq • ChromaDB + SQLite

AI-Powered E-Commerce Chatbot — RAG & Text-to-SQL

Semantic Query Routing, Vector Search & Natural-Language SQL Catalogue Discovery

Developed an LLM-powered e-commerce chatbot combining Retrieval-Augmented Generation (RAG) and Text-to-SQL for automated FAQ support and natural-language product discovery across brand, price, discount, and customer rating filters.

  • Implemented semantic query routing using MiniLM sentence embeddings to dynamically dispatch queries between FAQ RAG and Text-to-SQL pipelines.
  • Engineered FAQ retrieval with ChromaDB and ultra-low-latency response generation powered by Llama 3.3 through the Groq API.
  • Built an automated Selenium web scraping and data ingestion pipeline populating a structured SQLite catalogue with 643 product records.
PythonLlama 3.3Groq APIChromaDBSentence TransformersSemantic RouterSQLiteStreamlitSeleniumpandas
Source Code
Vehicle Damage Classification Using Deep Learning
AI & Backend2025 — 2026
82.26% Validation Accuracy • 2,300 Labeled Images • FastAPI Inference

Vehicle Damage Classification Using Deep Learning

Fine-Tuned ResNet50 Computer Vision Pipeline Served via FastAPI & Streamlit

Developed a computer vision application capable of classifying vehicle damage images into six front and rear damage categories, coupling a PyTorch ResNet50 inference backend with a RESTful FastAPI service and interactive Streamlit frontend.

  • Trained and evaluated a fine-tuned ResNet50 convolutional neural network on 2,300 labeled vehicle images, achieving 82.26% validation accuracy.
  • Applied image augmentation, normalization pipelines, and automated hyperparameter optimization with Optuna.
  • Exposed real-time model predictions via a decoupled FastAPI REST endpoint consumed by a Streamlit image upload UI.
PythonPyTorchTorchvisionResNet50OptunaFastAPIStreamlitscikit-learn
Source Code
Global Superstore Data Warehouse & Business Intelligence
Data EngineeringFeb 2026 — May 2026
51,290-Row Dataset • SCD Type 1 & 2 • 5 Power BI Dashboards

Global Superstore Data Warehouse & Business Intelligence

51,290-Row SQL Server Star Schema, SSIS ETL Pipelines, SSAS OLAP Cube & Power BI

Designed an end-to-end enterprise SQL Server data warehouse and BI analytics solution for a 51,290-row global retail dataset using a five-dimension star schema and central sales fact table.

  • Developed automated SSIS ETL pipelines integrating SQL, CSV, and Excel sources with surrogate-key lookups, SCD Type 1 & Type 2 customer dimension updates, and quarantine handling for missing profit records.
  • Implemented transaction completion updates and automated order processing-time calculations.
  • Built a multi-dimensional SSAS OLAP cube with date, product, and geographic hierarchies alongside five interactive Power BI report pages for sales, profitability, shipping, and drill-down analysis.
Microsoft SQL ServerT-SQLSSIS (ETL)SSAS (OLAP Cube)Power BIStar SchemaExcel
Source Code
Let's Connect

Get in Touch

Currently seeking Site Reliability Engineering (SRE) and DevOps Engineering opportunities focused on production troubleshooting, operational automation, CI/CD, and cloud infrastructure.

Send a Direct Message

Have an SRE/DevOps role, cloud project, or university community collaboration in mind? Let's talk.