Youssef Ateya

My Experience

Posted on Oct 23, 2025

Senior Associate, Forward Deployed AI Software Engineer @ Kyndryl

February 2026 to Present · Dallas, TX

  • Full stack engineer shipping production GenAI applications and agentic systems in Python and TypeScript across AWS, Azure, and GCP on customer-facing engagements with large enterprise clients including American Express, a state DMV, and Dow, owning solution architecture, backend services, API contracts, and agent runtime.
  • Built an enterprise insights agent on GCP spanning 500+ applications, translating user intent, history, and active filters into structured JSON requests against a semantic-layer data-query agent, returning grounded narratives, tables, and charts with citations, rendered in the React/Next.js frontend I build alongside the Python backend.
  • Designed a multi-agent orchestration pipeline for a mainframe modernization program at a large enterprise customer, handling intent routing and agent chaining across discovery, process-mapping, notebook-builder, and validation agents in LangGraph, with JSON Schema contracts and validation gates between stages.
  • Built a source-code understanding pipeline that parses legacy COBOL and JCL into a queryable knowledge graph, with a custom parser handling copybook expansion and malformed source, persisted as a combined graph and vector store in MongoDB so engineers can search decades-old systems in natural language.
  • Contribute to Kyndryl Agent Builder, a company-wide GenAI platform with a capability registry and tool catalog so applications discover and invoke reusable skills at runtime, with tools exposed over MCP.

Digital Wildcatters Logo Software Engineer @ Digital Wildcatters

June 2024 to July 2025 · Houston, TX

  • Architected and deployed RAG-powered LLMs for enterprise energy clients, processing 100,000+ unstructured oil & gas documents across PDFs, Word, Excel, and PowerPoint using distributed microservices with Kafka.
  • Rebuilt ingestion around structure-aware semantic chunking, deduplication, and source normalization after fixed-size chunks destroyed tables and separated numbers from their row and column headers.
  • Built hybrid retrieval fusing Qdrant vector search over BGE embeddings with keyword search and reranking, so engineers could query exact well names and API numbers.
  • Engineered an end-to-end Azure ML pipeline in Python with Document Intelligence, Computer Vision, and Cognitive Search to automate ingestion and search, achieving 3× faster retrieval.
  • Fine-tuned an open-source Llama model in PyTorch on public oil & gas corpora.
  • Built full-stack features for Collide in TypeScript, React, and Ruby on Rails; containerized services with Docker and Kubernetes and held pytest/Jest/RSpec suites at 90% coverage via Jenkins and GitHub Actions.

Meta Logo Software Engineer Intern (MLH Fellowship) @ Meta

June 2024 to September 2024 · Remote

  • Optimized Clang/LLVM passes for 30% faster compilation in specific workloads, and published notes consisting of hundreds of pages of technical documentation for LLVM and Clangd, supporting 100+ new contributors.
  • Open source contributions to Clangd, Clang-Tidy, and ClangIR under the LLVM project, improving diagnostics, C++ code analysis, and Clangd Language Server features shipped to VS Code users.