Open to full-time remote roles
Birehan Zewdie, Full-Stack AI Engineer
shipping AI products end-to-end.
I build and ship production AI products end-to-end: RAG and LLM agents, FastAPI backends, and Next.js frontends. Recent wins: a 7% revenue lift from an AI floor-price model at The COOL Company, and a bilingual RAG legal platform live in production.
4+ years of software / AI engineering. LLMs, RAG & AI agents · LLM fine-tuning & evaluation · FastAPI & Python · Next.js, React & TypeScript · PostgreSQL / pgvector · Docker on GCP & AWS.
01: Selected work
Production systems that shipped
End-to-end AI products, owned from frontend and backend through RAG, agents, and cloud deploy. Shipped and running for real users.
02: Experience
Where I've shipped impact
Production ML and full-stack delivery across ad-tech, health AI, and advertiser tooling: modeling, LLM pipelines, and React dashboards.
AI/ML Engineer @ The COOL Company
May 2025 to PresentNew York, USA (Remote)
- ›7% revenue increase from an AI floor-price prediction system (LightGBM) that optimizes pricing per ad request for DSPs.
- ›86%+ accuracy on a client revenue-forecasting system built from scraped domain data, PageSpeed and WHOIS signals, historical performance, and LLM-generated features.
PythonLightGBMLLMsAI IntegrationFeature EngineeringDSP / Programmatic AdsWeb Scraping & SignalsModel DeploymentMachine Learning Engineer @ Adludio
May 2024 to May 2025London, UK (Remote)
- ›~16% KPI lift from an automated pacing algorithm that pairs predictive modeling with real-time bid adjustments.
- ›~10% engagement lift and ~5% CTR/CPA improvement from a creative scoring system (historical performance + LLMs) that ranks variants before spend.
- ›Cut creative ideation time roughly in half with an LLM pipeline that generates on-brand ad concepts.
- ›~15% conversion lift and ~10% margin lift from an inventory scoring model (DCN v2 + learning-to-rank) that tunes bids across publishers, formats, and devices.
PythonPyTorchDCN v2LTRLLMsAI Software Engineer · Team Lead @ Eskalate S.C.
Mar 2022 to Apr 2024Addis Ababa, Ethiopia
- ›Led a cross-functional team of 18 to launch HakimHub, a first-of-its-kind AI-powered medical recommendation platform, coordinating planning and delivery through Jira.
- ›Fine-tuned a Llama 3-based chatbot using AI agents for personalized symptom assessments and follow-ups, integrated with tooling to query doctors and hospitals.
Llama 3AI AgentsPythonNext.js
03: Skills
What I work with every day
The tools I reach for first, by layer. Primary stack is highlighted, and everything here has shipped to production.
AI / ML Engineering
PrimaryBackend
PrimaryCloud · DevOps
PrimaryFrontend / Full-Stack
PrimaryAutomation · Integrations
04: Side projects & research
ML experiments and open-source research
LLM fine-tuning, RAG architectures, and ML pipelines. Most include a Medium write-up and GitHub source.
Legal Contract Advisor: High-Precision RAG for Legal Q&A
Feb 2024
Contract Q&A RAG system for Lizzy AI built to deliver high-precision answers on legal documents using semantic chunking, hybrid retrieval, and a fully evaluated RAG pipeline.
Redash Chatbot Add-on: Agentic RAG for Data Analysis
2024
Conversational chatbot add-on for Redash that translates natural-language questions into SQL/queries against existing dashboards using an agentic RAG approach.
Loan Risk Prediction ML Pipeline (LightGBM + FLAML + MLflow)
Sep 2024
End-to-end loan risk prediction pipeline: data ingestion, preprocessing, LightGBM training, FLAML hyperparameter tuning, MLflow tracking, Docker, and CI/CD retraining via GitHub Actions.
Fine-tuning Llama 2 for Amharic Text Generation
2024
Fine-tuned Llama 2 to enable quality embeddings and text generation in Amharic, then used it inside a RAG-based ad-copy builder for the Ethiopian market.
Automated Storyboard Synthesis for Digital Advertising
Feb 2024
ML solution that automated storyboard creation for digital ads by combining EDA on creative assets, YOLO object detection, and image generation.
n8n Content Pipeline: Topic to Published SEO Article in Under 4 Minutes
2025
Self-hosted n8n pipeline triggered by one Telegram command. Six stages: parallel research, SEO analysis, a multi-agent Writer/Critic/Refiner loop, Claude QA, and final scoring.
05: Achievements & education
Recognition and academic record
Awards & honors
10 Academy Cohort A, Valedictorian (with Distinction)
Top of leaderboard in a 6-month intensive Machine Learning, Generative AI, Data Engineering, and Web3 program. Less than 4% of applicants completed.
2nd Place, A2SV Champions League
Competed against 500+ students; champions-league format with 32 finalists.
View3rd Place, Collegiate Programming Contest (EtCPC)
Out of 80+ teams from universities across the region.
ViewTop Problem Solver, A2SV G4 Camp II
Solved every camp coding challenge; 1st place top problem-solver certificate.
Winner, DevFest 2022 Hackathon
Hosted by Google Developers Group.
Education
Oct 2019 to Jun 2024
AAU
BSc in Software Engineering (AI Stream)
OOP, Databases, Software Development, Machine Learning, NLP, Reinforcement Learning.
Dec 2023 to May 2024
10 Academy
Data Science · ML · Generative AI · Web3 (Distinction, Valedictorian)
Sep 2022 to Sep 2023
A2SV (Africa to Silicon Valley)
Software Engineering Program (backed by Google)
Solved 1,000+ algorithm problems (850+ LeetCode, 200+ Codeforces).
06: About
Engineer first, AI second.
I'm Birehan, a full-stack AI engineer with 4+ years of experience building production products where AI is wired in end-to-end: Next.js frontends, async FastAPI backends, and the RAG pipelines and LLM agents behind them. I shipped Al-Shalawi, a bilingual Arabic/English legal platform, from database and API through the RTL UI and a RAG drafting assistant. Before that I built pricing and revenue tooling at The COOL Company, and an LLM pipeline at Adludio that cut concept iteration time roughly in half.
I own the whole slice: frontend, backend, model, and the evaluation glue that makes teams trust what ships. That sits on a Software Engineering degree (AI stream), 1,000+ competitive problems through A2SV, and valedictorian at 10 Academy Cohort A.
07: Contact
Hiring or have a project? Let's talk.
Email or WhatsApp both work. I reply within a business day; WhatsApp is fine for a quick ping.
Open to full-time remote Full-Stack AI / AI Software Engineer roles./in/birehan
@birehan
Medium (@birehan)
+251982070195
Email to schedule