Arteria AI

We are excited to be exhibiting at InvestOps USA 2025 from March 10-12! Want to learn more about Arteria? 👋 Swing by Booth #9 to chat with Samantha Lippitt, Donna Mansfield, Jeremy A. Siegel, and Christine Abanto. 🎤 Come hear Samantha Lippitt discuss 'Managing Cross Functional Risk' alongside other industry leaders on March 11th at 3:20 pm. See you there! #investops #investmentoperations #AI #Data #Fintech

We are thrilled that our very own Samantha Lippitt will be speaking at InvestOps 2025 - stay tuned for more updates on how you can connect with Arteria AI at the conference. #InvestOps #AI #Fintech

Absolutely could not be prouder of legendary Arterians Amir Hajian, Shima Khoshraftar and Niaz Abedini for publishing their latest research: GraphiT: Efficient Node Classification on Text-Attributed Graphs with Prompt Optimized LLMs. cc. Arteria AI, Abrar Huq, Jonathan Wong, Jason Wild, John Wallace, Information Venture Partners, Citi, BDC, Notable Capital, Bailey Dougherty, Mark Bennewith, Brent Taylor, Achille Ettorre, Eric Etu, Andy MacDonald, Mahtab (Maria) Kamali, PhD, Alexander Ross, Mark Beeston, Jeff Richards, Chelcie Taylor, Jen Holmstrom, Robert Antoniades, Guy Cogan, Catherine Y. Zhai   #AI #artificialintelligence #research #cafe #arteria

🚨 New Research Alert!   We're thrilled to share our latest research, GraphiT: Efficient Node Classification on Text-Attributed Graphs with Prompt Optimized LLMs. GraphiT combines multiple flavours of AI (graphs and LLMs) and will be presented at ACM Web Conference 2025.   The Café is perked to share its work! Swipe through the carousel to get the highlights and be sure to check out the full paper in the comments below! #AI #Research #Fintech #Innovation

🚨 Excited to share our most recent research: 🚨 GraphiT: Efficient Node Classification on Text-Attributed Graphs with Prompt Optimized LLMs 💡 The main idea is to leverage large language models to tackle node classification tasks by optimizing graph encodings and the LLM prompt. The paper will be presented at the Non-Euclidean Foundation Models and Geometric Learning (NEGEL) workshop of TheWebConf 2025. Paper link in the comments. ✅ Summary: ➡️ Encoded the information in the graph structure into text by picking the optimal attributes and including a proxy for the summary of the neighborhood (see below) and neighbor labels (where available)  ➡️ Used semantic key-phrases instead of neighborhood summaries. This resulted in a noticeably smaller input context length without a loss of critical information (smaller context length = less expensive LLM cost). For a 1-hop neighbor classification, using keyphrases brought down the token numbers by a factor of 5 on average.  ➡️ Used automatic prompt optimization instead of manual prompt tweaking. This led to a significant improvement in the performance of the classifier (~3% to 9% improvement). ⚡ What’s next: ➡️ There is still a gap between a pure LLM approach and a hybrid LLM/GNN approach, which we think can be closed using a few ideas that we are experimenting with.  ➡️ The idea of auto-prompt optimization combined with key-phrase extraction can be used in other graph-ML tasks. The underlying auto-prompt optimization is based on the powerful DSPy framework that we regularly use in our LLM projects at Arteria AI. Working on this with Shima and Niaz has been a lot of fun, and I am looking forward to more of these collaborations.

We’re excited about the launch of our advanced AI research lab, the Arteria AI Café!☕ Today, we’re sharing the full interview, with all the latest episodes in one place. Join our Chief Science Officer, Dr. Amir Hajian and CEO, Shelby Austin as they chat about what’s brewing at Arteria AI and how our clients can sample the Café's newest innovations! #AI #Fintech #Innovation

Arteria AI's own Samantha Lippitt will be speaking at and meeting with the buy side community at InvestOps USA 2025 in Orlando next month! Let us know if you'll be there. #ArteriaAI #FinTech #DocumentationInfrastructure #BuySide

We're delighted to serve our customers within Arteria's newest AI research arm, the Café.   In this fourth episode, our Chief Science Officer, Dr. Amir Hajian and CEO, Shelby Austin, talk about the 'secret sauce' to building with LLM's!☕ #AI #Fintech #Innovation

Happy National Latte Day! At Arteria AI, we believe that great ideas – like great coffee – require the right environment to flourish. The Café is that environment: a space where our team can experiment, innovate, and explore the big questions, one "shot" of inspiration at a time. Stay tuned for the final episode of our interview series this week and a full version next week! #brewingideas #ai #innovation

We're thrilled to share more details about Arteria's newest AI research arm, the Café. We have some ideas percolating - specifically on Agentic AI☕   In this third episode, our Chief Science Officer, Dr. Amir Hajian and CEO, Shelby Austin, dive into agentic knowledge systems and its implications on software. #AI #Innovation #Fintech