1. Scope
Pick the exact job: answer from documents, draft reports, recommend actions, automate steps, or support admins.
AI SaaS features people can use.
I build RAG assistants, embeddings/vector search, document ingestion, AI reports, recommender flows, and the SaaS layer around them.
Useful AI, connected to real data, shipped with auth, billing, audit logs, tests, and deployment.
Value
One owner across product, frontend, backend, AI, and release.
How
Short scope, AI-assisted build loops, direct updates, and targeted tests.
Outcome
A working flow users can test, teams can sell, and engineers can maintain.
Packages
Clear offers with a concrete shipped result.
AI MVP Sprint
A testable AI feature: RAG chat, vector search, document Q&A, report generator, recommender, or agent workflow.
SaaS Feature Sprint
The SaaS layer around the feature: onboarding, auth, roles, billing, admin views, API, database, and release.
AI Product Polish
Make an AI flow more useful: embeddings, retrieval quality, prompts, streaming UX, empty states, errors, and trust signals.
Shipping Hardening
Multi-provider routing, cost tracking, audit logs, AI governance checks, eval cases, tests, CI/CD, and handover.
What you get
The practical upside of working with me.
Concrete AI output: answers, recommendations, generated reports, completed tasks, or admin decisions.
AI connected to product data, documents, embeddings, permissions, users, billing, and workflows.
Measurable delivery: fewer manual steps, faster support/admin work, better retrieval quality, shorter time to MVP, or clearer conversion flow.
Cleaner handover with prompts, tests, decisions, and code another engineer can continue from.
Proof
Relevant work, not filler.
Data Wipe Mailer
A simple, privacy-focused web app that helps Swedish citizens exercise their GDPR Article 17 right to erasure by sending deletion requests to Swedish data brokers like MrKoll, Ratsit, and others. Built with Next.js and hosted on Vercel, operating entirely client-side with no server data transmission. My LinkedIn post about this project received 2k+ likes and 200k+ views with overwhelmingly positive feedback.
TypeScriptNext.jsTailwind CSSshadcn/ui
48
9
J.A.R.V.I.S
A private, local AI assistant that integrates with personal notes, tools, and workflows. Built with SwiftUI for macOS, using Ollama for local LLM execution and MCP servers for system tool integration like iCloud Notes and SL Stockholm public transit. Runs entirely on my machine with no cloud dependencies.
SwiftUIOllamaMCP
0
0
Carspotter
Carspotter is a social media platform for car enthusiasts built by me and a group of students for the course "Project in Software Engineering" at Stockholm University, Spring term 2023. Users can upload images of cars they've seen, keep track of them on a map, earn badges/achievements and look at profiles and the cars they've seen on a grid.
FlutterGoogle Maps APIPlate Recognizer APIFirebase
0
1
How delivery works
Short, direct, and built for delivery.
2. Build
Connect ingestion, embeddings, retrieval, model calls, UI, permissions, logging, fallbacks, and the SaaS flow.
3. Ship
Deploy it, test real cases, check cost and failure modes, then hand over the next useful iteration.
Send a short brief.
Send the product, scope, timeline, and the measurable outcome you need. I will reply with what I would ship first and why.
Project details
Opens a pre-filled email draft to work@joelhagvall.com.
Include enough detail to explain the scope, timeline, and goal.
Copy the brief first if you want to save it before opening mail.
Direct contact
Good briefs include the user, the blocked flow, the deadline, and what should be true after the work is done.
Value
One owner across product, frontend, backend, AI, and release.
How
Short scope, AI-assisted build loops, direct updates, and targeted tests.