[go: up one dir, main page]

AI Engineer

I build AI products around real problems, not just model demos.

I am Cyprian Kiplangat, an AI Engineer from Kenya working across language technology, agents, robotics-flavored interfaces, and fast product experiments. Most of my work lives in the space between research curiosity and practical software that people can actually try.

Priority project

Swahili - Kalenjin Translator

A low-resource language translation project focused on making local AI more useful for indigenous Kenyan languages.

A lot of AI progress still misses indigenous and low-resource languages, which means communities like mine are left behind by systems that claim to be universal. I started this project to push against that gap by fine-tuning NLLB 600M locally on an 8GB VRAM laptop, iterating for hours at a time, and turning it into a Swahili-Kalenjin translation system with public models, a research preview, and room to grow into something much bigger.

  • This is the project I want people to see first because it sits closest to the kind of future I want to build.
  • It combines language access, constrained-hardware experimentation, and a practical view of how AI can matter locally.
  • The work already lives across public models, a live preview, and a demo, even though the training code stayed local.
AINLPLow-Resource LanguagesResearch

2025 - now / Long-term research and product direction

Selected work

The rest of the portfolio branches out from that same instinct: use AI where it meaningfully changes the product.

Some projects here are polished demos, some are hackathon builds, and some are experiments that mattered more for the idea than the finish. I still want them visible because they show how I think.

2026Browser demo

SafeArm AI

A browser-based robotic arm simulation that you control through natural language.

AIRoboticsSimulationHackathon
2026Shipped demo

RepoX

A learning tool that turns any GitHub repository into an interactive explanation and exploration experience.

AIDeveloper ToolsEducationVisualizationHackathon
2026Shipped demo

Persite

A personalization demo that changes messaging based on locale and visitor intent without leaning on invasive tracking.

AIPersonalizationLocalizationWebHackathon

About

I care most about building AI work that feels useful, grounded, and worth returning to.

I study Information Technology at Kabarak University, and I tend to approach software with one question in mind: can this idea become something a real person can use, trust, or learn from? That is why a lot of my work sits at the intersection of language, interfaces, systems, and product thinking instead of staying as model experiments alone.

Professionally

I like applied AI most when it is attached to a clear problem: local language access, better developer learning, more intuitive interfaces, or tools that make complex systems easier to work with. I am comfortable using AI to move faster while building, but I still judge the work by whether the final product feels coherent and useful.

Personally

Outside code, I am pulled toward science and esoteric questions, especially the huge ones about where we came from, why we are here, and where we might be going. I like hearing people talk about quantum physics, subatomic particles, and string theory even when I only understand part of it. The Big Bang Theory is still one of my favorite comfort shows, and The Matrix keeps staying with me because it feels like philosophy and technology asking the same old question about reality from different directions. That curiosity probably explains a lot about how I think and build.

Tools and stack

The stack changes with the problem, but these are the tools that show up across the work most often.

Most of these projects sit somewhere between product building, model integration, and experimentation, so the toolbox is broader than one framework or one AI provider.

Languages and core

PythonTypeScriptJavaScriptNode.jsBun

Frontend and product

ReactNext.jsTanStack StartTailwind CSSViteD3.jsBabylon.jsThree.js

AI and model work

Hugging FacePyTorchGeminiCerebrasNLLBLLM agentsMCPVisionAgents

Infra and deployment

Cloudflare WorkersCloudflare PagesVercelWranglerR2