Hi, my name is

Kidane Gebremedhin

Senior Full-Stack & AI Engineer | I design distributed, highly available systems, and the AI agents that run on top of them.

Seven years architecting fault-tolerant systems in FinTech, Healthcare, and Education with Java, Spring Boot, Python, FastAPI, Node.js, React, and Next.js. These are platforms that move real money, onboard real customers, and cannot afford to be down. Today I bring that same discipline to AI products: RAG pipelines, agentic workflows built with LangGraph and LangChain, and multi-tenant LLM services that stay grounded, observable, and affordable under production traffic. I treat an AI feature as a distributed system with a probabilistic component, not a demo, which means hard tenant isolation, evaluation before shipping, and sane behaviour when the model or the vendor has a bad day. I like hard problems, and I care most about the ones where good engineering shows up on the business ledger.

About Me

I am a senior full-stack engineer with 7 years of experience designing enterprise systems in FinTech, Healthcare, and Education, the kind where an outage means lost originations, a delayed diagnosis, or a payment that cannot be reconciled. My work spans high-throughput lending and onboarding platforms, event-driven retail systems, financial rails integrations, and clinical record systems, built with Java, Spring Boot, Python, FastAPI, PHP, Node.js, React, and Next.js. I optimize for systems that survive contact with production and stay cheap to maintain years later.

Increasingly that work is AI-shaped: retrieval-augmented generation, vector search, agent orchestration with LangGraph and LangChain, tool-calling and MCP integrations, and multi-tenant LLM platforms serving real customers. The engineering problems that decide whether an AI product succeeds are rarely the model. They are retrieval quality, tenant isolation, evaluation, latency, and cost per conversation. That is the layer I work at: grounding answers so the system can be trusted in front of customers, isolating tenants so it can be sold to businesses, and instrumenting everything so regressions are caught by telemetry rather than by users. Underneath it all I am still the engineer who enjoys profiling a slow query, drawing service boundaries that hold up, and designing APIs other teams are glad to build on.

My Skills

AI Engineering

  • Agents & orchestration: AI agents, LangGraph, LangChain, multi-step workflows, tool calling, MCP, human-in-the-loop
  • RAG & retrieval: Pinecone, embeddings, chunking strategies, hybrid search, re-ranking, grounding & citations
  • LLMOps: prompt engineering, evaluation harnesses, guardrails, tracing, token & cost optimization
  • Providers & tooling: OpenAI, Anthropic, OpenRouter, Firecrawl, OpenAI-compatible APIs

Backend Development

  • Languages: Java, Python, Scala, PHP, JavaScript
  • Frameworks: Spring Boot, FastAPI, Play Framework, Laravel, Node.js, Yii, NextJS
  • Concepts: OOP, Design patterns, Microservices, gRPC, RESTful APIs, Event Driven architecture

Databases & Data

  • Relational: PostgreSQL, MySQL
  • NoSQL: MongoDB, Firebase
  • Data: SQL, PL/pgSQL, Snowflake, Sumologic

DevOps & Cloud

  • Cloud: AWS, Vercel, Supabase
  • Containers: Docker, Kubernetes
  • CI/CD: GitHub Actions, Jenkins

Frontend and Mobile

  • Frameworks: NextJS, React, Vue.js, Flutter, Android
  • Languages: JavaScript, TypeScript, Dart, Java/Kotlin
  • Styling: Tailwind CSS, CSS3

Collaboration & Communication Tools

  • Collaboration: Git, Jira, Confuluence, Lucid Chart
  • Communication: Slack, Discord, Zoom, Google Meet

Projects & Case Studies

AI Customer Service Chatbot

A multi-tenant AI support platform that lets any business drop a chat widget onto its site and deflect repetitive support conversations to an agent grounded in its own knowledge base. Operators get a real-time inbox, human handoff, analytics, and self-serve billing, so the product onboards and monetizes customers without engineering in the loop. Retrieval is namespaced per organization and agent, making tenant isolation a hard architectural guarantee rather than a policy promise.

TypeScriptNext.jsReactExpressMongoDBRedisSocket.ioRAGPineconeOpenRouterFirecrawlNextAuthPaddleMinIOTurborepoTailwind CSSDockerGitHub Actions

Distributed Retail Platform

An event-driven retail platform that keeps inventory and order state accurate in real time across millions of requests, so the business stops overselling stock it does not have and customers see the truth at checkout. Decomposing the domain into Spring Boot microservices let teams ship independently instead of queueing behind a single release train, and contained failures so a degraded payments or fulfillment path never takes the storefront down with it.

JavaSpring BootPythonFastAPIMicroservicesPostgreSQLMongoDBKafkaRedisGrafanaDockerKubernetesAWSGitReactNextJSREST APIsgRPC

Universal Onboarding

A customer onboarding platform that turns a cold signup into a funded borrower with no human in the loop: registration, credit application, KYC submission and evaluation, and disbursement are all automated end to end. Every manual step removed is a step applicants no longer drop out of, so the architecture is built around conversion as much as throughput. It sustains thousands of concurrent users and high transaction volume, because during an acquisition push, onboarding downtime is lost originations that never come back.

JavaSpring BootRedisPythonFastAPIPostgreSQLScalaPlay FrameworkKafkaDockerKubernetesAWSGitAndroidReactNextJSREST APIsgRPC

Rails provider API integration for credit service

Integrated Coins.ph as a financial rails provider so customers could move money in and out of the platform directly, unlocking a payment channel the business could not previously reach and removing a manual reconciliation burden from operations. Money movement is unforgiving, so the pipeline was built security-first and reconciliation-first: every deposit and withdrawal is idempotent, auditable, and provably settled against the provider.

JavaSpring bootRedisKafkaLaravelPythonFastAPIScalaPlay FrameworkWebSocketRESTful APIGit

Electronic Medical Record (EMR)

An Electronic Medical Record system that gives clinicians one current view of a patient at the point of care: medical and treatment history, medications, and test results in a single place, instead of scattered across paper files and disconnected systems. Faster access to accurate records means less time chasing charts, fewer avoidable errors from missing history, and more of the visit spent on the patient. Built to be highly customizable so each provider can model its own clinical workflows without a separate fork of the product.

JavaSpring bootPythonFastAPILaravelYIIMySQLGitAWSFlutterVueNextJSREST APIsgRPCRedis

Get in Touch

I am always open to discussing new projects, hard technical problems, or collaboration opportunities, especially work that puts AI in front of real users and has to hold up. Available for senior full-stack and AI engineering roles, as well as freelance engagements: greenfield AI products, RAG and agent systems, or scaling an existing platform that has outgrown its architecture.