Software engineer combining 7+ years of production software experience with ML expertise. I build scalable systems that actually ship to production.

• Building end-to-end ML pipelines with PyTorch and deploying them at scale
• Contributing to open-source ML projects
• Exploring transformer architectures and their applications in NLP and computer vision

Machine Learning & Data

• Frameworks: PyTorch, TensorFlow, scikit-learn
• MLOps: MLflow, Docker, FastAPI
• Data: pandas, NumPy, Hugging Face

Software Engineering

• Languages: JavaScript/TypeScript, Python, PHP, Node.js
• Frontend: React, Next.js, Vue.js
• Backend: Node.js, GraphQL, REST APIs
• DevOps: Docker, AWS, GitHub Actions, CircleCI
• Databases: PostgreSQL, MongoDB, Elasticsearch

Currently Senior Software Engineer at X-Team, previously delivered high-impact features at scale for startups and enterprises accross news media, education and travel industries.

Skillset:

• Production systems at scale (millions of users)
• Complex system migrations and architecture
• Cross-functional collaboration with non-technical stakeholders
• Strong testing, CI/CD and reliability practices

AI-powered real estate rehab cost estimation using computer vision and multi-modal LLM. A production-ready system that analyzes property photos to estimate renovation costs, combining vision-language models with domain-specific cost modeling. Integrates as a Chrome extension for Zillow listings.

• Multi-modal LLM (Google Gemini Vision) for image understanding
• Automated feature extraction: property condition, material quality, damage assessment
• Multi-image analysis for comprehensive property evaluation (interior, exterior, structural systems)
• Confidence scoring based on image quality, coverage, and AI certainty

• Rules-based estimation engine with 20+ repair categories
• Feature engineering: size-based multipliers, market adjustments, regional factors
• Ensemble approach: combines AI analysis with domain heuristics
• Output: itemized cost breakdown with uncertainty quantification

Handling variable image quality and angles from real estate listings Extracting structured data from unstructured vision model outputs Calibrating AI confidence scores to real-world accuracy Dealing with incomplete information (not all property areas photographed)

• LinkedIn: linkedin.com/in/jasperkinoti
• Email: jasper dot kinoti at gmail