Scroll
About Me
I'm an AI Software Engineer with 2+ years of experience building production-ready Generative AI systems and resilient backend services on AWS, currently based in Colombo, Sri Lanka.
I specialize in designing RAG pipelines, orchestrating LLM workflows using LangChain & LangGraph, and deploying scalable REST APIs with Python (FastAPI). I'm passionate about bridging complex AI capabilities with real-world enterprise applications.
Currently pursuing a BSc in Information Systems at the University of Colombo School of Computing, with hands-on experience in AI safety evaluation, MLOps frameworks, and cloud-native infrastructure.
0
Years Experience
0
Projects Completed
0
Certifications
Skills & Expertise
GenAI & LLM Engineering
RAG Pipelines
LangChain & LangGraph
Prompt Engineering
LLM Integration (OpenAI)
Backend & Cloud
FastAPI & REST APIs
AWS (EC2, S3, Lambda)
Docker & CI/CD
MLOps & AI Safety
Technologies
Python
FastAPI
LangChain
LangGraph
OpenAI API
Hugging Face
AWS
Docker
MongoDB
MySQL
Pandas
Git
Featured Projects
1 of 4
KYC-Vision | Identity Verification Pipeline
Automated document processing pipeline using Python, Tesseract OCR & OpenCV to digitize scanned National Identity Cards. Implements Regex & Named Entity Recognition to parse unstructured OCR output into structured JSON for identity verification.
LLM Data Studio | AI Intent Extraction
AI prototype using Transformers (DistilBERT) for aspect-based sentiment analysis — extracting structured business insights from unstructured user text with multi-label classification and loss-function optimization.
Automated Supply Chain ETL Pipeline
End-to-end data pipeline orchestrated with Apache Airflow (Docker) for automated daily processing. Implements data extraction from AWS S3, transformation with Python/Pandas, quality checks, and structured loading to support downstream analytics and ML workloads.
TrackNBook | Real-Time Logistics System
Real-time logistics and scheduling automation for the Sri Lankan railway network. Features a concurrent seat reservation engine, automated warrant processing, and an FSM-inspired asset tracking module with automated conflict resolution logic.
Publications
1 of 4
Towards AI
Small Language Models (SLMs): A Practical Guide to Architecture and Deployment
A comprehensive breakdown of SLM architecture — covering Sparse Attention, Knowledge Distillation, LoRA fine-tuning, and real-world edge deployment scenarios. Featured in Towards AI with 54 claps.
Towards AI
QwQ-32B: Alibaba's Tiny Titan Challenges AI Giants in Reasoning
Deep technical analysis of Alibaba's QwQ-32B reasoning model — benchmarked against DeepSeek-R1, OpenAI o1, and Claude 3.5 Sonnet. Covers RL training, efficiency gains, and real-world deployment considerations.
Medium
DeepSeek Database Exposure: A Wake-Up Call for AI Security
Technical breakdown of the January 2025 DeepSeek ClickHouse database breach — covering the misconfiguration, exposed data types, implications for AI infrastructure security, and actionable lessons for the industry.
Medium
Deep Learning 101: Why It's Changing the Game (Part 1 & 2)
A two-part beginner-to-intermediate guide demystifying deep learning — covering neural network architectures (CNNs, RNNs, LSTMs, Transformers, GANs), activation functions, loss functions, and optimization algorithms in plain English.
Experience
Machine Learning Engineer
Synapse AI Labs
December 2024 – Present
Designing and deploying LLM-powered applications and RAG pipelines using LangChain and vector embeddings. Building scalable AI-serving architectures on AWS (Lambda, EC2) with FastAPI. Contributing to AI safety evaluation — implementing hallucination detection metrics and MLOps practices in an Agile enterprise environment.
Backend Developer (Remote)
Liulian Tech Sdn Bhd — Malaysia
December 2023 – December 2024
Built robust RESTful APIs using Python (FastAPI) integrating backend services with frontend platforms and external data ecosystems. Managed cloud infrastructure on AWS, implemented monitoring and alerting systems, and optimized async database logic for high-traffic applications.
Research Assistant
UCSC Research Projects
April 2023 – December 2023
Validated NLP datasets for low-resource languages (Tamil/English) to support emotion detection models. Built data preprocessing scripts to automate the cleaning and annotation of 1,000+ text entries and video datasets for computer vision training.
Get In Touch
Interested in collaborating or have a project in mind? Feel free to reach out!