Francisco De Haro

Full Stack Developer · Next.js · AI & ML

Contact

Email

frandeharo55@gmail.com

Location

Spain

Portfolio

deharo.dev

Profile

Results-oriented full-stack developer. Experience in 0→1, fast cycles, and end-to-end products. Specialized in Next.js, React Native, and AI. Published Repai Trainer on App Store with subscriptions, Supabase backend, and LLM integration. I prefer clear code, rapid experimentation, and products people want to use.

Availability

Open to freelance projects, technical collaborations, or positions where I can build product from day one.

Summary

Full-stack developer focused on Next.js/TypeScript and Python, shipping user-centric products end-to-end (web + mobile with Expo) with solid backends (Supabase/PostgreSQL/Node). Strong with APIs—designing and integrating REST/GraphQL/streaming endpoints and third-party SDKs (auth, payments, observability). Practical AI integrations (Mistral, OpenAI, Llama): prompt orchestration, evals, and lightweight RAG; shipped products with subscriptions and CI/CD. Immediate availability; open to relocation and a sabbatical year for the right project.

What I bring

  • Frontend (Next.js/React) and mobile (Expo) design and development
  • APIs and backend in Node/TypeScript (Supabase, PostgreSQL, Neon)
  • Authentication and payments (Clerk, Stripe, RevenueCat, subscriptions)
  • LLM integration (Mistral, OpenAI, etc.), prompt orchestration and deployments
  • Model fine-tuning (Llama, QLoRA, LoRA) and image generation (Stable Diffusion)
  • Practical DevOps: Vercel, Cloudflare, GitHub, CI/CD

Projects

Promecenas

B2B Fintech Infrastructure for NGOs

Live
  • Stripe Connect & Split Payments: Implementation of connected accounts
  • Multi-tenant Management: Database architecture in Supabase
  • Webhooks & Synchronization: Robust payment event listening system
  • Dynamic Pricing Algorithm (Tiered Pricing) for commissions
  • Retention Engine (Churn Prevention) for subscriptions
Next.jsStripe ConnectSupabasePostgreSQLTailwind CSSFintech

Canivos (Now Promecenas)

B2B Fintech Infrastructure for NGOs

Live
  • Stripe Connect & Split Payments: Implementation of connected accounts (Managed Accounts)
  • Multi-tenant Management: Database architecture in Supabase
  • Webhooks & Synchronization: Robust payment event listening system
  • Dynamic Pricing Algorithm (Tiered Pricing) for commissions
  • Retention Engine (Churn Prevention) for subscriptions
Next.jsStripe ConnectSupabasePostgreSQLTailwind CSSFintech

Repai Trainer

iOS Mobile App Development (Fullstack)

Live
  • Personalized workout generation with OpenAI API
  • Subscription system (monthly, annual, lifetime) with RevenueCat
  • Exercise database integrated with Gymvisual
  • Progress tracking and statistics
  • Complete user management with Supabase Auth
ExpoReact NativeSupabasePostgreSQLRevenueCatApp Store

UMChat - Advanced Chat System

Intelligent web chat application with document management, complete authentication and complex architecture

Live
  • Document management system (Vault) with hierarchical organization
  • Automatic OCR with Mathpix API for PDFs
  • Intelligent file deduplication with SHA-256
  • Complete authentication with Supabase Auth (OAuth, email/password)
  • Guest mode with limited functionalities
Next.jsReactTypeScriptxAI APISupabasePostgreSQL

Testum

AI-powered test exam platform

Live
  • Creation and management of personalized multiple-choice exams
  • Integrated AI chat with OpenAI for assistance while answering
  • User system with Supabase Auth
  • PostgreSQL database to store exams and answers
  • Edge Functions for optimal performance
Next.jsOpenAISupabasePostgreSQLEdge FunctionsTypeScript

Demo Mistral Chat

Chat interface with Mistral AI and streaming

Live
  • Real-time response streaming
  • Modern and responsive chat interface
  • Conversation context handling
  • Edge Functions for minimal latency
  • Prompt optimization for better responses
Next.jsMistral AITypeScriptEdge FunctionsStreaming

Llama Fine-tuning for Spanish Lyrics

Fine-tuning Llama 3.3 and 3.1 models for music lyrics generation

Archived
  • Fine-tuning Llama 3.3 (1B/3B) and 3.1 (8B) with QLoRA 4-bit
  • Custom dataset of +10,000 prompt-completion examples
  • Autonomous scraping pipeline with Playwright for capture and parsing
  • Intelligent tokenization with masking (labels=-100 in prompts)
  • Optimized training with Unsloth (~2× faster)
PyTorchLlamaFine-tuningQLoRAHugging FaceUnsloth

LoRA Black Stories - Image Generation

LoRA training to replicate Black Stories aesthetic

Archived
  • LoRA adapter training on FLUX (Black Forest Labs)
  • Base models: black-forest-labs/flux-schnell and flux-dev
  • Curated dataset with autocaptioning for consistency
  • Trigger word system ("TOK") for prompt-based control
  • Multi-resolution training (512-1024px)
FLUXLoRAComfyUIImage GenerationFine-tuningComputer Vision

1¢, 1min, 100%: AI in University Exams (In Development)

Opensource research on AI models applied to Business Administration exams

Dev
  • Opensource research: results and methodology publicly available
  • Models evaluated: DeepSeek API, xAI API, OpenAI API, Google AI API
  • OCR system implemented with Mathpix API for exam digitization
  • Automatic exam conversion program to LATEX/Markdown format
  • SQLite database with Prisma for exam and results management
ResearchAIPythonOCRMathpixLATEX

Tech Stack

Frontend & Mobile

Next.jsReactTypeScriptExpoReact NativeXcodeTailwind CSS

Backend & Database

Node.jsTypeScriptSupabaseNeonPostgreSQL

Auth & Payments

ClerkStripeRevenueCatApp Store Connect

AI & Machine Learning

LLMsMistral AIOpenAIPyTorchFine-tuningLoRAQLoRAStable DiffusionHugging FacePrompting

DevOps & Tools

VercelCloudflareGitHubCI/CDCursor

Education

Business Administration and Management Degree (BBA)

Currently 3rd year

Fast learning and technical autonomy

Languages

Spanish

Native

English

Advanced

French

Basic

Available immediately for collaborations and new opportunities.

Contact → frandeharo55@gmail.com