yuna@yunadevefy.me:~$ whoami
I build AI products people actually use.
> ai_ml_engineer_and_frontend_lead · addis_ababa
0+
years shipping production software
0K+
monthly users on AI features I built
0%
ML query latency cut through backend redesign
0K+
monthly actives on TravelColombia.ai
00 / terminal
Talk to the site.
welcome to yunadevefy.me
type 'help' and hit enter to look around
01 / what i do
The full stack of an AI product.
RAG and agentic systems
Retrieval pipelines and multi-step agent workflows that hold up in production, built with LangChain, LangGraph, FastAPI, vLLM, and Qdrant.
AI-native frontends
Streaming interfaces that make model output feel instant, from SignalR token streams to optimistic UI, in React, Next.js, and TypeScript.
NLP for low-resource languages
Custom hate-speech and misinformation detection models for Amharic and other Ethiopian languages, a space where off-the-shelf models fail.
02 / the pipeline
How an AI feature ships.
The same path runs under everything I build, from TravelColombia.ai to the logistics RAG engine at Adilas Tech. Data in, grounded answer out, streamed into an interface that keeps up.
03 / selected work
Work that shipped.
04 / toolkit
The stack I reach for.
05 / career
Where I have been.
Nov 2025 - Present
AI and Frontend Lead · Senior AI/ML Engineer
Adilas Tech
Leading AI development on a logistics platform built around an end-to-end streaming RAG application.
- Architected the full path from a React frontend over SignalR to a high-throughput FastAPI inference service running vLLM.
- Built asynchronous retrieval against a Qdrant vector store so operators asking questions in plain language get answers grounded in the platform's own data.
- Designed the streaming protocol to render tokens the moment they exist, keeping perceived latency low even on long answers.
- Implemented a hybrid routing layer that decides per query whether a self-hosted model can answer or the request should escalate to GPT-4o, keeping costs predictable without capping quality on hard queries.
- Established the core patterns the platform uses for concurrent token generation.
Nov 2024 - Jul 2025
Senior Software Engineer
SkyHigh Cloud
Led frontend development across international client work.
- Delivered 10+ international client projects serving more than 100,000 combined monthly sessions.
- Integrated LLM features with the Gemini API and the Vercel AI SDK, lifting user engagement by 25 percent.
- Cut ML query latency by 40 percent by redesigning how the backend retrieved data.
May 2021 - Nov 2024
Senior Software Engineer · ML Engineer
AdotSense Digital
Built production NLP for languages that off-the-shelf moderation models do not cover.
- Trained TensorFlow text-classification models for hate-speech and misinformation detection in Amharic and other low-resource Ethiopian languages, where training data barely exists and commercial APIs offer no coverage.
- Owned the full lifecycle: data refinement from social media sources, training, evaluation, and production serving behind a Python API.
- Deployed the models inside an automated content analysis system running against live content.
- Built the React platform researchers used to explore model outputs.
- Drove a 170 percent traffic increase through analytics-led iteration.
Dec 2019 - May 2021
Founder · Software Engineer
Ma'ed Technologies
Founded a campus-focused startup and shipped its first two products.
- Launched a Flutter campus app covering navigation, real-time announcements, and the daily information students actually needed, reaching 2,000+ active users in its first year.
- Architected and built the frontend for a university procurement web application that moved a paper process of requests, approvals, and tracking online, cutting manual work by 65 percent.
- Grounded both products in 250+ user research sessions, gathering requirements directly from the people using them.
- The procurement application is still in continuous use.
06 / beyond the code
Community and credentials.
Community
Lead Organizer
Google Developer Group Adama
Organizing a regional developer community, running events and mentorship programs with a 90 percent project-completion rate among participants.
Mentor
ADPList and Microverse
Mentoring early-career engineers on career growth, code review habits, and breaking into AI engineering.
Credentials
- BSc in Computer Science and Engineering, Adama Science and Technology University
- AWS Generative AI certification
- Udacity AWS Cloud Developer Nanodegree
- Kifiya AI Mastery Program
07 / contact
yuna@yunadevefy.me:~$ start_a_project
Open to senior AI/ML and frontend engineering roles, remote or on-site, and to interesting collaborations. Based in Addis Ababa, Ethiopia.