Explore our robust AI-powered technology delivering powerful results from your complex data.
Request a Demo Today
RAG Pipeline for Unstructured Data
Advanced retrieval-augmented generation for enterprise knowledge management
Converts diverse data into uniform text chunks.
Creates searchable vectors in Elasticsearch.
Finds and prioritizes relevant information.
Produces verified answers from source material.
Research & Articles
An enterprise-friendly B2B AI platform that ensures data retention, privacy, and control
eSapiens' Derek Module: Deep Extraction & Reasoning Engine for Knowledge with LLMs
Journal: arXiv:2507.15863[cs.CL], Submitted on 13 Jul 2025, Author: Isaac Shi, Zeyuan Li, Fan Liu, Wenli Wang, Lewei He, Yang Yang, Tianyu Shi
In this paper, we present DEREK (Deep Extraction & Reasoning Engine for Knowledge), a secure and scalable Retrieval-Augmented Generation (RAG) pipeline purpose-built for enterprise document question answering. Developed by eSapiens, Derek processes heterogeneous content into optimized 1,000-token chunks, indexes them in a hybrid HNSW+BM25 store, and delivers answers via multi-stage query refinement, reranking, and CO-STAR–based prompt engineering. A LangGraph verifier enforces strict citation grounding, reducing unsupported statements to under 3%. Evaluated on LegalBench subsets, Derek achieved measurable gains in recall, precision, and factual grounding while operating in fully containerized, encrypted environments. Designed for high-stakes domains such as legal and finance, it offers accurate, auditable, and production-ready document QA with minimal operational overhead.
T2S Module for Structured Data
Natural language to SQL conversion for non-technical users
SELECT p.name, SUM(s.amount) AS total_amount FROM sales s JOIN products p ON s.product_id=p.id WHERE EXTRACT (YEAR FROM s.date)=2025 GROUP BY p.name ORDER BY SUM(s.amount) DESC
1
Task Analysis & Routing
Identifies structured query needs and routes accordingly.
2
SQL Construction & Execution
Builds read-only queries using schema knowledge.
3
Self-Correction & Rating
Fixes errors and improves low-quality queries.
4
Insight Generation
Transforms data results into natural language insights.
5
Answer Delivery
Presents findings in user-friendly format.
The Framework of Thor
Unleash the fury of development with the framework of Thor—where code strikes like lightning and scales like the hammer of the gods.
Request a Demo Today
Research & Articles
An enterprise-friendly B2B AI platform that ensures data retention, privacy, and control
Thor: Transformer Heuristics for On-Demand Retrieval
Journal: arXiv:2507.09592[cs.CL], Submitted on 13 Jul 2025, Author: Isaac Shi, Zeyuan Li, Fan Liu, Wenli Wang, Lewei He, Yang Yang, Tianyu Shi
In this paper, we introduce the THOR (Transformer Heuristics for On-Demand Retrieval) Module, designed and implemented by eSapiens, a secure, scalable engine that transforms natural-language questions into verified, read-only SQL analytics for enterprise databases. The Text-to-SQL module follows a decoupled orchestration/execution architecture: a Supervisor Agent routes queries, Schema Retrieval dynamically injects table and column metadata, and a SQL Generation Agent emits single-statement SELECT queries protected by a read-only guardrail. An integrated Self-Correction & Rating loop captures empty results, execution errors, or low-quality outputs and triggers up to five LLM-driven regeneration attempts. Finally, a Result Interpretation Agent produces concise, human-readable insights and hands raw rows to the Insight & Intelligence engine for visualization or forecasting.
Transform Your Business with AI-as-a-Service
Drive major success for your company with eSapiens.ai
