Joined January 2018
- Tweets 330
- Following 179
- Followers 177
- Likes 186
33 Photos and videos
Pinned Tweet
Jun 10
🚀 Bigger isn't better. A 24B open model just beat the giants.
Lettria's fine-tuned Mistral Small 3.2 to top the text-to-graph leaderboard, outperforming frontier systems over 15x its size.
🔗 perseus.lettria.com/leaderbo…
#AI #LLMs #OpenSource #FineTuning
1
1
63
Jun 4
🧠 AI is only as smart as its memory.
Most LLMs lack the structure for complex work. Enter Perseus by Lettria and turn unstructured data into an auditable Knowledge Graph.
🚀 Zero hallucinations
🚀 Fully auditable
🚀 Python SDK
Stop guessing. Start structuring.
1
31
May 28
Stop drowning in disconnected data. 🌊
"The Old Way": Every document is an island.
"The New Way": Perseus Interlink.
We read every variant across every document, turning 4 documents and 4 spellings into 1 unified graph.
🔗 perseus.lettria.com/
49
May 25
1/6 For the builders in the room:
The hardest part of #GraphRAG isn't the LLM, it’s the architectural bridge. You need a fast and reliable way to move from raw, messy files to a structured graph database.
Enter Perseus SDK: The Python-native toolkit for graph infrastructure
5
54
May 25
6/6
Stop building custom extractors and start building your graph. 🚀
The perseus client is open and ready for your next project. Dive into the code and the docs on GitHub:
🔗 github.com/Lettria/perseus-c…
#Python #SoftwareEngineering #GraphDB #Neo4j #FalkorDB #DataScience #Perseus
1
46
May 25
5/6
Why developers are switching to the Perseus SDK:
✅ Pythonic: Feels natural to any Data Scientist or AI Engineer.
✅ Non-blocking: Optimized for scale and high-volume workloads.
✅ Flexible: Manage ontologies, files, and graphs through a simple, unified interface.
14
May 25
4/6
Data integrity is non-negotiable. 🛡️
Perseus uses Pydantic for robust data validation. This ensures that every entity, relationship, and property traveling from your text to your graph matches your schema exactly. No more "ghost nodes" or corrupted data structures.
1
35
May 25
3/6
Native Integrations matter. 🔌
The SDK provides first-class support for the industry’s leading graph databases. With a simple pip install, you get direct connectors for:
🔹 Neo4j
🔹 FalkorDB
Zero custom boilerplate needed to start syncing your data.
14
May 25
2/6
Perseus was built for high-performance production environments. ⚡
It’s fully asynchronous (built on asyncio and aiohttp), meaning it handles high-concurrency ingestion and API calls without blocking your main event loop. Speed is a core feature, not an afterthought.
12
May 21
You are using multiple graphs? Now, you only need one. ☝️
Different spellings and disconnected documents shouldn't stall your insights. Perseus Interlink automatically resolves every variant in your data into a single unified graph.
#DataAI #KnowledgeGraph #TechTrends #Lettria
28
May 18
Building a Knowledge Graph is the right way to secure AI accuracy, but a nightmare for manual labor. 📉
The biggest bottleneck? Manual Ontology Modeling.
⏲Most teams spend weeks defining schemas before they even ingest a single document.
Here’s Perseus automates that👇
6
1
26
May 18
Explore the SDK:
🔗 perseus.lettria.com
#DataEngineering #KnowledgeGraph #Python #AI #GraphDB #Perseus
38
May 18
5/6 Stop building ontologies by hand. Let Perseus handle the architecture so you can focus on the insights.
The future of data engineering is automated. 🚀
10
May 18
4/6 Why developers love it:
✅ Zero-to-Graph in minutes: Skip the modeling phase.
✅Dynamic Adaptation: The schema grows as your data does.
✅ API-First: Integrate automated modeling directly into your CI/CD.
It’s the move from "Hand-coded" to "AI-orchestrated."
22
May 18
3/6 But data isn't static, and neither is Perseus.
It features automated versioning. As your data evolves, your schema evolves with it. No more manual migrations or broken pipelines when a new document type is introduced. 🛠️
17
May 18
2/6 Perseus flips the script. 🔄 It uses advanced agentic AI to automatically generate #ontologies directly from your unstructured data. It reads your corpus and identifies the natural structure, so you don't have to guess.
84
May 18
1/6
#DataEngineers, we know the pain:
1️⃣ Analyzing thousands of docs
2️⃣ Manually identifying entities & relationships
3️⃣ Mapping them to a rigid schema
It’s slow, it doesn't scale, and it breaks the moment your #data changes. 🛑
1
21
May 18
2/6 Perseus flips the script. 🔄
It uses advanced agentic AI to automatically generate #ontologies directly from your unstructured data. It reads your corpus and identifies the natural structure, so you don't have to guess.
93
May 13
RAG is the gold standard for AI, but it has a massive trust problem: Hallucinations.
🚫Traditional vector search often loses the context between entities, leading to "plausible-sounding" lies. It’s time to give your AI a real memory. A thread on how Perseus is fixing this. 👇
1
36
May 13
5/6 Perseus handles the heavy lifting:
✅ Automated entity extraction
✅ Relationship mapping
✅ Seamless integration with Neo4j & FalkorDB
✅ High-fidelity parsing of complex PDFs
It’s the infrastructure layer that makes "Enterprise-Grade AI" actually possible.
1
34
May 13
Ready to stop the hallucinations and start building high-precision AI? Check out the Perseus documentation and see how to bridge the gap between raw data and structured intelligence.
🔗 perseus.lettria.com
#GraphRAG #AI #KnowledgeGraph #LLM #DataEngineering
53