Aller au contenu principal
William Balance
FR EN KO

Freelance AI full-stack developer services

I work with SMBs, scale-ups and product teams that want to move from demo to product. Here are the five kinds of engagements I take end-to-end.

01

Custom AI applications

Domain chatbots, internal assistants, specialized agents with RAG. I design, build and ship full AI apps wired into your data and systems. Auth, permissions, observability, cost control. Stack: TypeScript, Next.js or Astro, Postgres, pgvector or Qdrant, Claude and GPT.

Typical outcomes
  • Internal assistant that answers on private docs with no data leak
  • Support chatbot deflecting 40 to 60 percent of tier-1 tickets
  • Agents with tools (search, internal APIs, actions) running in production
Pricing
On request
02

LLM integration in existing products

Adding AI features to a live SaaS or app: suggestions, generation, classification, extraction, semantic search. Streaming, retries, multi-provider fallback, caching, per-user budgets. I plug in OpenAI, Anthropic, Mistral or self-hosted open-source models depending on constraints.

Typical outcomes
  • New AI feature shipped in 2 to 6 weeks
  • API costs under control via caching and smart routing
  • Automatic fallback when a provider goes down
Pricing
On request
03

AI automation

Multi-step workflows that replace manual work: document parsing and classification, structured scraping, PDF and email extraction, cross-tool sync. Solid orchestration, queues, retries, readable logs. N8n or custom code depending on volume.

Typical outcomes
  • Document pipelines processing thousands of files per day
  • Smart scraping that adapts to structure changes
  • Several engineer-days saved per month on repetitive work
Pricing
On request
04

Classic full-stack development

When you just need a solid developer. SaaS, dashboards, APIs, client portals, back-offices. Modern stack: TypeScript, Next.js, Astro, Node, Postgres, Cloudflare. Clean deploys, simple CI, monitoring. No over-engineering.

Typical outcomes
  • MVP shipped in a few weeks, ready for real users
  • Real-time dashboards wired to your data
  • Documented and tested APIs
Pricing
On request
05

AI audit and strategic advisory

For SMBs that want to start without getting burned by the hype cycle. I review your processes, pick the use cases that actually pay off, and tell you what is worth doing and what is not. Deliverables: summary brief, prioritized roadmap, cost and ROI estimates.

Typical outcomes
  • 3 to 5 prioritized use cases with estimated ROI
  • Realistic 3 to 6 month roadmap
  • Data governance and security nailed down from day one
Pricing
On request

Got a project?

30 free minutes to scope your need and figure out whether I am the right fit.

Book a call [email protected]

Frequently asked questions

How long does it take to ship a project?

A typical LLM integration or new AI feature lands in production in 2 to 6 weeks. A full-stack MVP usually runs 4 to 10 weeks. I scope before I commit, and I would rather ship a useful V1 fast than a monster in six months.

Do you work with startups on a tight budget?

Yes, as long as the scope is realistic. I can shape an MVP that delivers value without burning a seed round. What I refuse: 50-feature plans priced like a junior's salary. We prioritize, ship, iterate.

How do you handle remote work from South Korea?

KST is +9h from Paris and +13h from New York. In practice: I cover EU hours in the morning and US or APAC in the evening. Regular calls, proper written docs, no ghosting. Many European clients actually appreciate it — work lands while they sleep.

Why are your rates on request and not public?

Because "integrate a chatbot" can mean 3 days or 3 months. I give a firm quote after a free 30-minute scoping call, with committed pricing and delivery date. No fake packages that balloon after signing.

Do you handle post-launch maintenance?

Yes, on request. I offer monthly maintenance contracts for the projects I ship: monitoring, fixes, small improvements, LLM model updates. No long commitment. If you want to take it in-house, the code is clean and documented for that handover.