Backend systems,data pipelines that ship

Python, FastAPI, pandas, NumPy, scikit-learn, Apache Airflow, Redis, Docker

A production-style movie recommendation service built on MovieLens interactions, item metadata and modular training and serving pipelines.

• Combined content similarity, item co-occurrence and popularity signals to support stable recommendations for sparse user histories.
• Exposed FastAPI endpoints for health checks, personalized recommendations, similar-item lookup and model refreshes.
• Separated ingestion, feature engineering, training, evaluation and serving layers, with Airflow support for scheduled workflows.
• Added Docker Compose, optional Redis caching, API-key protection, structured logging, request tracing and CI-ready tests.

Python, TensorFlow, TensorFlow Lite Micro, Flask, AWS

A final-year research project exploring lightweight sensor-data compression and reconstruction for edge-to-cloud communication.

• Developed an autoencoder model to compress and reconstruct sensor data before cloud-side processing.
• Built preprocessing and training pipelines in Python and TensorFlow, tuning latent-vector size for compact representation.
• Converted the encoder for microcontroller-side inference and exposed the decoder through a Flask API for cloud reconstruction.
• Deployed the backend API on AWS EC2 with S3 integration for encoded payload handling.

AWS Lambda, S3, SQL, Python, analytics workflows

An automated ingestion and transformation pipeline for moving raw operational data into analytics-ready datasets.

• Automated ingestion with AWS Lambda and S3 to reduce manual handoffs and keep data delivery consistent.
• Standardized raw inputs through transformation jobs that improved downstream usability.
• Delivered SQL-based analytics views for customer activity and operational reporting.

WebRTC, REST APIs, Angular integrations, backend services

Voice and video communication SDK work focused on more reliable integrations across frontend, backend and third-party systems.

• Improved integration reliability for real-time voice and video experiences.
• Built REST APIs that connected frontend clients, SDK modules and third-party platforms.
• Worked across Angular, backend and SDK teams to ship end-to-end product features.

AWS EC2, Docker, IAM, VPC, S3, ALB, rate limiting

Backend APIs and cloud services supporting approximately 1,000-3,000 daily users through AWS-backed infrastructure.

• Deployed containerized backend services on AWS EC2 with IAM, VPC and S3 integrations.
• Configured production API delivery behind AWS Application Load Balancer with rate limiting and traffic controls.
• Improved platform security through time-bound access controls and abuse-resistance measures.
• Implemented a chatbot workflow to reduce first-response friction for customer interactions.

Apache Kafka, PySpark Structured Streaming, Apache Airflow, Postgres, Docker, S3-compatible storage

A medallion-architecture (bronze/silver/gold) pipeline for e-commerce order analytics, taking simulated orders from a Kafka topic through to a Postgres star schema.

• Streamed Kafka-published orders into a bronze Parquet layer on S3-compatible object storage with PySpark Structured Streaming.
• Refined raw events through batch Spark jobs into silver, then modelled a gold star schema of customer, product and date dimensions around a sales fact table.
• Orchestrated the batch stage with an Airflow DAG, running Kafka and Airflow on Docker Compose.
• Documented the architecture, environment contract and known limitations, including the hardcoded Hadoop path and the manually started streaming processes.