Backend and Data Engineer with a background in Computer Engineering and an MSc in Advanced Computer Science from the University of Dundee.

I work primarily on backend systems, data engineering, and applied AI.

My work includes:

• Backend APIs and services
• High-volume ETL / ELT pipelines
• Data cleaning and entity resolution
• Database architecture and optimization
• AI and LLM integrations
• Retrieval and knowledge systems
• Developer tooling and automation

Some production systems listed below were built in commercial environments, so their source code is not publicly available.

Lead backend engineer for a multi-tenant AI gateway providing a single OpenAI-compatible API across 13 AI providers, including OpenAI, Anthropic, DeepSeek, xAI, fal, and ElevenLabs.

• Built provider adapters across 13 AI services
• Implemented weighted API-key rotation and automatic failover for 429, 402, and 401 responses
• Built asynchronous media-processing workflows
• Designed batched request logging and analytics pipelines using ClickHouse
• Implemented per-request usage and cost calculation
• Added per-client spending limits and usage controls
• Built CSV/XLSX reporting exports
• Served 6M+ requests
• Reduced monthly request failure rate from approximately 42% to ~1%
• Responsible for approximately 92% of repository commits

Stack: Elixir · ClickHouse · PostgreSQL · API Integration · ETL · AI APIs

Built the core data platform for a product-price comparison system combining multiple retailers and marketplaces with a large government product database.

• Processed roughly 100M raw product records
• Consolidated them into approximately 5M canonical products
• Used ClickHouse for high-performance cleaning, joins, and transformations
• Designed hybrid entity resolution using GTIN matching and source-specific deduplication
• Built custom matching logic for sources where standard identifiers were unavailable
• Loaded the consolidated dataset into ArangoDB for indexed retrieval
• Designed the backend API serving the application

Stack: Python · ClickHouse · SQL · ArangoDB · ETL · Web Scraping · Entity Resolution · API Design

Built and evaluated multiple NLP approaches for argument relation classification.

• TF-IDF + Linear SVM
• BERT embeddings + SVM
• Fine-tuned RoBERTa
• Flan-T5
• XGBoost
• Dockerized prediction API
• Batch and single-item inference endpoints

Stack: Python · PyTorch · Hugging Face Transformers · scikit-learn · Flask · Docker

Built a reverse-image search backend using neural image embeddings and similarity-based retrieval.

• Generated image embeddings using a CNN-based model
• Implemented similarity-based image retrieval
• Built the backend API with FastAPI
• Used background workers for asynchronous processing
• Stored application and processing data in PostgreSQL

Stack: Python · FastAPI · PyTorch · PostgreSQL · Celery · Docker