BuildingIntelligent Systems

Hi, I am ThilankaAI & Full Stack Developer

I build AI-driven applications, production-minded ML pipelines, and data-intensive systems that turn research ideas into usable software.

Portrait of Thilanka Wijesingha

Expert Technologies

Data Engineering

Distributed and batch/stream pipelines

  • Apache Kafka
  • Apache Airflow
  • Apache Hadoop
  • YARN
  • Apache Spark
  • PySpark
  • Power BI

AI and ML

Model workflows, orchestration, and observability

  • Python
  • LangChain
  • LALangGraph
  • LangFuse
  • MLflow

Cloud and DevOps

Cloud delivery and engineering workflows

  • AWS
  • Docker
  • Git

Backend and APIs

Service architecture and app backends

  • Node.js
  • Express.js
  • FastAPI
  • Flask
  • Java
  • Playwright

Databases

Operational and vector data platforms

  • PostgreSQL
  • MySQL
  • Microsoft SQL Server
  • MongoDB
  • Neo4j
  • Firebase
  • Qdrant
  • Supabase

Web and Mobile

Frontend and client experience

  • Flutter
  • React
  • Next.js
  • JavaScript
  • HTML
  • CSS
  • Tailwind CSS

Work Experience

Sep 2025 - Aug 2026

Software Engineer

Contract
Standard Deviant Analytics
  • Developed a scalable, multi-module Spring Boot data platform that ingests and processes data from 300+ external game providers.
  • Built provider-specific ingestion modules, reducing the average integration time for new providers to under two days.
  • Maintained automated, multi-threaded data workflows with 98%+ reliability while processing millions of records.
  • Designed a slot-game automation system covering spin target selection, screenshot capture, ROI configuration, PaddleOCR extraction, normalization, and validation.
  • Automated the extraction of Balance, Bet, Win, and spin-type data from large volumes of gameplay screenshots, reducing manual review and improving consistency.
  • Reworked a 3+ hour-per-game OCR process into a configuration-driven parallel pipeline that runs multiple capture and extraction jobs while preserving per-game sequence accuracy and isolated outputs.

Oct 2024 - Apr 2025

Full Stack Engineer Intern

Internship
Nextzela Technologies
  • Built cross-platform applications with Flutter for responsive user interfaces
  • Launched Next.js web solutions with focus on performance and UX
  • Designed MySQL databases to support efficient querying
  • Ensured reliable application behavior across all platforms

Featured Projects

Kapruka Agent system architecture showing FastAPI, LangGraph orchestration, specialist agents, memory layers, Qdrant retrieval, Supabase, crawling pipelines, Tavily search, and Langfuse observability.

Kapruka Multi-Agent Gift Concierge

Stateful multi-agent gift concierge that combines product retrieval, customer memory, logistics reasoning, semantic caching, corrective retrieval, and observable LangGraph orchestration.

PythonFastAPILangGraph+9
View
Repo
Telco churn prediction system architecture showing the production-style machine learning workflow.

Telco Customer Churn Prediction Pipeline

Production-oriented churn modeling workflow with PySpark preprocessing, classical ML baselines, MLflow tracking, and streaming-style inference telemetry.

PythonMLflowscikit-learn+3
View
Repo
Hotzy Foods homepage hero with sauce bottles, navigation, and product discovery call to action.

Hotzy Foods Full-Stack Commerce Platform

Full-stack Next.js commerce and admin platform for Hotzy Foods, with MongoDB, JWT auth, RBAC, Cloudinary media, cart, checkout, orders, offers, and operational dashboards.

Next.jsReactTypeScript+7
View
LiveRepo
Default portfolio cover image for the MindMirror Personality Forecasting Engine case study.

MindMirror Personality Forecasting Engine

Async Big Five direction prediction service for weekly behavioral trend analysis from Reddit activity with tracked model lineage and authenticated API delivery.

Python 3.11FastAPIPostgreSQL+5
View
Repo
Default portfolio cover image for the MindMirror Conversational AI Agent case study.

MindMirror Conversational AI Agent

Domain-restricted conversational intelligence service that interprets psychology signals, personal history, and behavioral trends through a guarded LangGraph pipeline.

Python 3.10FastAPILangGraph+5
View
Repo
Default portfolio cover image for the Prime Lands Real Estate Intelligence Platform case study.

Prime Lands Real Estate Intelligence Platform

Config-driven RAG platform that transforms JavaScript-rendered real estate content into a grounded question-answering knowledge base with adaptive retrieval.

PythonLangChainQdrant+3
View
Repo

Build With Clarity

EmailLinkedInGitHub

Work With Me

I design and ship production-grade AI systems with reliable data foundations, strong delivery discipline, and measurable business outcomes.

Best Fit Projects

  • Machine Learning Systems
  • LLM Applications and Agent Workflows
  • RAG Architectures
  • Data Engineering Platforms
  • Full-Stack Projects

Availability

Available for full-time opportunities and freelance engagements, with flexibility for remote collaboration across time zones.

Let's Talk...