J·C·Fjelstul
Consulting LLC

Work

End-to-end ML,
from problem to production.

Selected Work
Hybrid Retrieval for Legal Tech

LawLibrary.ai

Architected a multi-stage hybrid recommender system commercializing IUROPA research. Engineered a domain-adapted bilingual BERT model with graph-augmented retrieval to achieve a 45% increase in Recall@10 against baseline embeddings.

Core Tech
Bilingual BERT / HyDE / Cross-Encoders
Evaluation
600k+ QA Pairs / 30k Ground-Truth
Role
Co-Founder & Principal Data Scientist
Unsupervised Semantic Mapping

Acuis.ai

Legal intelligence platform extracting structured insights from 60k+ EU competition law cases. Implemented a pipeline using UMAP and HDBScan on fine-tuned embeddings to cluster 17k insights into semantic threads with automated labeling.

Clustering
130 Automated Semantic Threads
Infrastructure
FastAPI / Supabase / Postgres
Data
60k+ Cases / 40k+ Decisions
Legal Database Infrastructure

The IUROPA Project

Project lead for an international research collaboration developing a comprehensive database on the CJEU. Architected a 20+ table relational system, a 500M+ text corpus, and a text-to-SQL RAG app.

Database
20+ tables
Output
500M+ Word Bilingual Corpus
System
Text-to-SQL RAG
Project
Agentic Sports Intelligence

WorldCups.ai

Developed an agentic text-to-SQL system that translates natural language queries into complex SQL against historical sports data. Built a commercial-grade API with rate-limiting and observability for public data access.

Pattern
LLM-to-SQL RAG Architecture
Engagement
230+ Stars on GitHub
Stack
Python / FastAPI / Next.js
Project
Ready to build?

From Problem Framing to Production.

Whether you need a domain-adapted text classification model, or an end-to-end recommender system with RAG, I help you ask the right questions, frame your business problem, and build cutting-edge AI/ML solutions.

Schedule a consultation