Bemnet Aschalew

Built 10+ LLM-powered automations in n8n integrating various tools, lifting outbound outreach volume by 70%+. Developed authenticated MCP servers for Google Analytics, Search Console, and Framer.

Built and deployed a RAG chatbot using Google ADK and AG-UI protocol, cutting LLM token consumption by 80%. Led migration of client platform from React.js to Next.js with server-side rendering and CI/CD pipeline.

Selected for a competitive software engineering fellowship focused on advanced DSA. Solved 700+ problems on LeetCode and Codeforces with strong placement in cohort contests.

Built a customer-facing AI agent from scratch using Google's ADK, shipping a working MVP in a 4-week timeline. Designed a serverless backend on Google Cloud with FastAPI and Pydantic RESTful APIs.

Built a LangGraph multi-agent system for an internal Agentic SDLC platform. Implemented an automated IEEE-standard documentation generator and engineered a Redis-backed memory persistence layer.

Selected among 24 fellows for a 10-day intensive program in computer vision and generative models. Implemented and trained custom diffusion models in PyTorch and authored a research write-up.

Activities and societies: Solved 700+ problems on LeetCode & Codeforces, with strong placement in cohort Codeforces contests.

Selected for a competitive software engineering fellowship focused on advanced data structures and algorithms. Solved over 700 problems on LeetCode and Codeforces, with strong placement in cohort contests.

Grade: 3.9/4.0

Activities and societies: AI Stream Project: Trained and deployed a custom CNN for image classification served via Flask.

Pursuing a B.Sc. in Software Engineering with an AI Stream, expected to graduate in July 2026. Achieved a cumulative GPA of 3.9/4.0 and developed a custom CNN for image classification served via Flask.

Activities and societies: Implemented and trained custom diffusion models in PyTorch focusing on temporal and class conditioning; investigated spatial and semantic control in 2D generative pipelines.

Selected among 24 fellows for a 10-day intensive program in computer vision and generative models, organized by Cornell-affiliated researchers. Implemented and trained custom diffusion models in PyTorch and authored a research write-up on spatial and semantic control in 2D diffusion models.