@erikapat.bsky.social
Inside Boston AI Week 2025: human-centered AI, skills, and scaling from a region packed with universities & startups. I caught part of it—here are the takeaways and why these events still matter (even if you can’t attend every session): medium.com/womenintechn..., #BostonAIWeek
Inside Boston AI Week — 2025
What mattered: human-centered AI, skills, and scale
medium.com
October 16, 2025 at 4:54 PM
Inside Boston AI Week 2025: human-centered AI, skills, and scaling from a region packed with universities & startups. I caught part of it—here are the takeaways and why these events still matter (even if you can’t attend every session): medium.com/womenintechn..., #BostonAIWeek
Reposted
LLMs don’t fail in the lab... they fail in production. This new article by Erika G. Gonçalves walks you through what really breaks in AI apps: silent retrieval failures, prompt drift, hallucinations, biased outputs + more.
AI Operations Under the Hood: Challenges and Best Practices | Towards Data Science
Building robust, reproducible, and reliable GenAI applications requires a framework of continuous improvement, rigorous evaluation, and systematic validation
towardsdatascience.com
September 5, 2025 at 8:15 PM
LLMs don’t fail in the lab... they fail in production. This new article by Erika G. Gonçalves walks you through what really breaks in AI apps: silent retrieval failures, prompt drift, hallucinations, biased outputs + more.
Reposted
Many AI models promise high accuracy but fail in real-world scenarios. Often data leakage is to blame. Erika G. Gonçalves explains how adopting a time-aware graph approach can prevent future-peeking and ensure reliable fraud detection.
No Peeking Ahead: Time-Aware Graph Fraud Detection | Towards Data Science
How to implement leak-free graph fraud detection
towardsdatascience.com
September 15, 2025 at 9:19 PM
Many AI models promise high accuracy but fail in real-world scenarios. Often data leakage is to blame. Erika G. Gonçalves explains how adopting a time-aware graph approach can prevent future-peeking and ensure reliable fraud detection.