Theo Korir
Full-Stack Engineer & Web Designer
New to Contra
Theo is ready for their next project!
A vibrant, mobile-first website for Fun Fiesta Kenya, a professional party planning service for children in Nairobi.
The site needed to feel fun and trustworthy at the same time — energetic enough to excite kids, polished enough to convert parents who are making a real purchasing decision.
What I built:
Bright, playful UI designed around the Fun Fiesta brand identity
Mobile-first layout built for Nairobi parents browsing on their phones
Clear service sections to communicate packages and offerings at a glance
Fast-loading, SEO-optimized structure for local discoverability in Nairobi
Integrated contact flow to capture party booking inquiries directly
Stack: React, TypeScript, Framer
Result: A production-ready website that turns first-time visitors into party booking inquiries — built for a local Nairobi audience.
1
12
Got it — a portfolio case study for the Serene Manor project. Here it is:
Serene Manor — Luxury Real Estate Website
A premium, fully responsive marketing website for Serene Manor, a luxury mixed-use development in Kilimani, Nairobi.
The brief required a site that felt as high-end as the property itself — elegant typography, smooth animations, and a layout that guides potential buyers from first impression to contact form without friction.
What I built:
Full multi-section site covering the hero, units, amenities, payment plans, gallery, and contact
Smooth scroll animations using Framer Motion for a premium feel
Mobile-first responsive design that holds up on every screen size
SEO-optimized semantic HTML structure for organic discoverability
Modular component architecture for easy content updates
Stack: Next.js 15 · Tailwind CSS · Framer Motion · TypeScript · Lucide React
Result: A production-ready site that positions Serene Manor as a premium brand and captures leads directly through an integrated contact section.
1
16
A Python-based automation tool that converts logo images into machine-ready DST embroidery files — eliminating the need for expensive digitizing software or manual tracing.
The pipeline uses OpenCV for contour and color detection, SciPy for smooth Bezier curve generation, and the OpenAI API for intelligent pre-analysis of complex or ambiguous logo regions. It correctly handles inner letter holes, fine text details, and multi-color separation — outputting a .DST file ready for industrial embroidery machines with zero manual intervention.
Built for businesses, print shops, and apparel brands that need to scale logo digitization without scaling headcount.
Tech used: Python · OpenCV · OpenAI API · NumPy · SciPy · Pillow
0
27
Built an end-to-end automation pipeline in Python running on Android via Termux on resource-constrained hardware
▸ Pipeline: link inbox monitoring → video download → ffmpeg frame extraction → parallel caption generation via
OpenAI Vision API → clipboard output
▸ Developed companion Android app (Kotlin) with a system share sheet integration — users share TikTok links directly
into the queue from any Android app
▸ Hardened server startup with automatic port conflict resolution using lsof/ss/netstat/fuser
▸ Demonstrated ability to deploy ML-adjacent tooling on non-standard hardware with zero cloud infrastructure
1
41
▸ Fine-tuning Microsoft WavLM Base (a self-supervised speech model) for emotion recognition using PyTorch and
HuggingFace Transformers
▸ Managing full training pipeline: data loading, augmentation, evaluation metrics, and checkpoint management
▸ Applying transfer learning best practices to adapt a large pre-trained model to a domain-specific objective
1
40
▸ Curating and cleaning a Kenya-specific audio corpus for DAPT of a speech/language model
▸ Applying data quality filters, deduplication, and preprocessing pipelines to ensure training data integrity
1
37