01
Tariff8AI - AI-Powered Customs Classification
No more manual customs code lookup. AI finds the right code in seconds.
ImpactClassification reduced from hours to seconds
PythonFastAPILlamaIndex+2
Loading...
Loading...
Guillaume Schonrock builds RAG systems and business chatbots in Python that run on your own infrastructure — no documents sent to third-party services. For startups and SMBs across Europe. 50,000+ documents searchable in under 100ms.
Response within 24h. No sales follow-up if it's not the right time.
Your team spends time hunting for information across shared drives, poorly organized wikis, and archived emails. The answer exists somewhere — but nobody knows where. You redo work, give clients wrong answers, make decisions on incomplete information.
SaaS tools like Notion AI or ChatGPT Enterprise promise to fix this — but they require you to upload your documents to their servers. For sensitive data (contracts, client records, IP), that's a non-starter.
A custom RAG chatbot queries your document base in real time, generates accurate, sourced answers, and runs entirely on your infrastructure. Your data never leaves your perimeter.
I build RAG (Retrieval-Augmented Generation) pipelines with LlamaIndex or LangChain, indexed in a Qdrant or pgvector store, connected to your existing document sources (SharePoint, Google Drive, database, internal API). The LLM can be OpenAI (via your API key) or an open-source model hosted on your infrastructure.
Document base of 50,000+ technical files with no way to query it
RAG pipeline with semantic embeddings and Qdrant vector index
50,000+ documents searchable in plain language in under 100ms
Sensitive document classification requiring a sovereign solution
Local LLM (Llama / Mistral) + classification pipeline fine-tuned on your categories
92% accuracy on proprietary data, zero dependency on external SaaS
Extracting key information from thousands of heterogeneous documents
LangChain / AutoGen agents with extraction and cross-validation tools
−90% manual data entry on document workflows
No more manual customs code lookup. AI finds the right code in seconds.
Ask a question, find the answer. Even across 50,000 documents.
AI reads your documents and extracts key information. No more manual entry.
I start by deeply understanding business goals and user needs.
Architecture decisions that support growth and maintainability.
Clean code, comprehensive testing, and thorough documentation.
Focus on measurable outcomes and real business impact.
A generic chatbot (ChatGPT, Gemini) answers from its public training data. A RAG chatbot queries your document base in real time — answers are precise, traceable, and sourced directly from your own documents.
The system is deployed on your infrastructure or a private server you control. Your documents are never sent to a third-party service. If you use OpenAI via your API key, only the queries (not the full documents) pass through their servers.
There's no theoretical limit. Production projects index 50,000+ technical documents. Search performance stays under 100ms even at scale thanks to optimized vector indexes.
Web app, automation, or AI integration — let's talk about how I can help.
Response within 24h. No sales follow-up if it's not the right time.