Vitali Avagyan

His Imperial doctoral thesis examines electricity-sector investment and profitability under uncertain fuel prices, demand and renewable generation. It connects financing constraints with the risk of a company’s cash flows and studies how energy storage profits move with other generators’ profits. Imperial records a PhD award in June 2017.[4]

An earlier American University of Armenia seminar biography records study in mathematical economics and operations research at Armenian State University of Economics and identifies him as an AUA alumnus. His current professional accounts link that research background with AI engineering and his work at Anterior.[1][3][5]

Use code relationships as retrieval context

Code-Graph-RAG parses a repository into functions, classes, modules and their relationships. A language model turns a question into a graph query, retrieves matching code and can propose edits with a diff preview. Tree-sitter supplies the common structure, while optional compiler frontends and runtime traces add facts that static parsing alone may miss.[2]

Keep the analysis local to the project’s capabilities

The default graph is an embedded SQLite file, with Memgraph and Neo4j available as alternatives. Runtime traces record calls that actually ran, not every possible path. Compiler-derived facts depend on the available toolchain. These boundaries matter when using the graph to investigate references, dead code or a proposed change.[2]