Thomas Assalian
BSc Computer Science — Concordia University, Montreal
Work Experience
2025
- Developed a scalable agentic RAG application, achieving 23× faster response times and $200,000 annual compute savings by optimizing retrieval pipelines, caching strategies, and model orchestration.
- Engineered advanced Retrieval-Augmented Generation (RAG) pipelines in Python by using LangChain and Google Vertex AI, improving document query relevance by 52%.
- Automated ingestion of 10,000+ internal files on Google Cloud Platform by chunking and embedding into MongoDB with MySQL-backed metadata, streamlining RAG document retrieval.
- Built and deployed 15+ task-specific Agents in Python by using LangGraph and LangChain, orchestrating REST APIs and internal tools via FastAPI to enable dynamic, goal-driven workflows.
- Refactored the AI agent SDK API for end-to-end async execution of agent tool calls; built a custom async HTTP client for REST APIs, improving tool execution speed by up to 40%.
- Boosted test coverage to 98%+ by implementing unit and integration tests across AI agent and RAG services.
- Applied Object-Oriented Programming to RAG and Agent architectures, enabling scalable orchestration, cleaner API integration, and more maintainable code.
- Containerized all services with Docker and deployed across Kubernetes clusters, achieving 99.9% up time.
- Tested and documented RESTful APIs using Postman and Swagger to validate proper functionality.
2024
- Engineered Python ETL pipelines (Pandas/NumPy) to extract, clean, and structure operational datasets powering Tableau dashboards, improving efficiency by 80.2% and reducing manual reporting work by 95%.
- Delivered 100% data accuracy for 30+ stakeholders by building automated validation + reconciliation scripts and partnering with business analysts to ensure reliable, decision-ready reporting.
- Verified datasets used in ML workflows, ensuring 100% compliance with AI project standards for accuracy and relevance
- Improved data quality by 20% by automating integrity checks and standardizing labeling + validation processes across multiple projects.