Skip to content
View moses-y's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report moses-y

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
moses-y/README.md

Hi, I am Moses Yebei

AI platform and site reliability engineer. I build AI systems that run in production, and I keep them running.

Portfolio LinkedIn Medium Book a call


About me

  • Nairobi, Kenya. Eight years in software, data and AI engineering.
  • I build document intelligence, LLM agents and the platforms under them: Kubernetes, infrastructure as code, observability and on-call.
  • I measure what I ship: accuracy, latency, cost and error rates for each component.
  • I wrote the preprint "Context Handoff Intelligence for AI Agents". It shows that agents lose context when they hand work to other agents, and that nobody benchmarks this.

What I work on

Project What it does Result
Trade finance document AI (a major African bank) Extracts and validates letter of credit documents, runs sanctions screening and Eximbills checks. A reviewer makes every decision. Self-hosted 27B model on one 48GB GPU. Cost model at 1M documents per year settled the build or buy decision.
Agent platform (fast-paced US startup) Autonomous repository maintenance with 7 parallel agents on a knowledge graph (149 entity types). I led AI engineering and ran production reliability on GKE.
Tender intelligence pipeline Crawls procurement portals in 12 countries and UN portals, then scores and routes notices with an LLM-as-a-judge gate. Runs on a schedule with monitoring.
Invoice automation Finance and audit automation with approvals and audit logs. Cycle cut from about 15 days to under 1 hour.
Legal document intelligence (top US law firms) NLP and ML pipelines for legal document review. 1M+ documents at 99 percent accuracy.

Tools I use

GitHub activity

GitHub streak


Open to AI Platform Engineer, Site Reliability Engineer and LLM Engineer roles. Remote or Nairobi.

Popular repositories Loading

  1. Data-Annotation-for-Beginners Data-Annotation-for-Beginners Public

    Data Annotation for Beginners: A Guide to Understanding and Automating the Process Using Python.

    4

  2. ycombinator-job-scraper ycombinator-job-scraper Public

    Y Combinator Job Scraper This repository houses an automated job scraping tool designed to streamline the job search process for tech professionals. Focused on Y Combinator's job board, this projec…

    Python 4 1

  3. YOLOv8_Segmentation_DeepSORT_TRACKING_SpeedEstimation YOLOv8_Segmentation_DeepSORT_TRACKING_SpeedEstimation Public

    Estimating speed using YoloV8 , Google Colab by modifying predict.py Script in segmentation folder.

    Jupyter Notebook 2

  4. qwiklabs_challenges qwiklabs_challenges Public

    Forked from GDSC-IIIT-Kalyani/qwiklabs_challenges

    Qwiklabs challenges helper guide

    1

  5. moses-y.github.io moses-y.github.io Public

    portfolio

    HTML 1 2

  6. ecommerce-chatbot ecommerce-chatbot Public

    This repo contains an Agentic Chatbot (Ari) to query a Database and get Order Status Delivery , Request to speak with a Customer Care Agent or ask about Return. All done using Google Gemini AI.

    Python 1