Amedea Pharma

Introducing New Centralized Performance and Strategy Intelligence in one Platform - ANCORA. ...Have you ever met a colleague in life sciences who was excited about capturing their performance after a busy, long day? Neither have we at Amedea. ...But in 2025, that has to change — because we cannot grow, we cannot lead, and we cannot even survive if we don’t know where to spend our time to replicate our personal bests—or even set new world records at work. So we are connecting the purpose of capturing your performance in Medical Teams to live, realistic views of the progress you are making and rewards for your contributions and collaborations with teams as well as your customers. Set up a meeting to schedule an ANCORA demo to drive your team's ultimate outcome and collaborations to the next level in 2025: https://lnkd.in/eJxH8rBP #medicalaffairs #performancekpis

Even when stakes are high and room for error seems lower than ever, our Age of Information calls for Transparency in Competition and Respect for each other through Sportsmanship to uncover more Winners in Medical Affairs across our industry. Watch this 1-minute highlight from Jeff New, our Expert Panelist during the last monthly webinar from Amedea Pharma a company that is bringing back excitement and purpose to this vital function in pharma and biotech via its new Performance and Strategy Intelligence solution called ANCORA! Sign up now for the next webinar March 20 @ 10:00 am ET by clicking here to receive slides + bonus content: https://lnkd.in/epwZCH4c featuring 2 special guests and pharma executives Sandra Walker and Eddie Power who will join our panelists Luz Andrea Ramirez Alex Pasek, MD! We will be speaking about an intriguing topic of innovative ways Industry must partner via Medical Affairs with HCPs to directly improve healthcare quality! Link to the full on-demand replay of the February webinar: https://lnkd.in/edHXaAJb #medicalaffairs #performancekpis #healthcarequalityimprovement

As you use new solutions to drive more outcomes think about being more intentional about everything you do and the real journey it takes to drive those patient and business results that matter in pharma. Here is a highlight from our last monthly webinar at Amedea Pharma from Luz Andrea Ramirez Watch the on-demand recording here: https://lnkd.in/eqikUWVN and contact me to today explore our new Operational Intelligence platform called ANCORA. Speakers: Luz Andrea Ramirez Jeff New Alex Pasek, MD Next webinar: MAR 20 @ 10 am ET - stay tuned for more details #medicalaffairs #performancekpis #aihealthcare

Sign up here https://lnkd.in/eCMEaCKX now to discuss how to add excitement and meaning to analyzing our performance in Medical Affairs #performancekpis #medicalaffairsimpact

Here's everything you need to know about Elon and xAI's launch of Grok-3, their new, state-of-the-art AI model.

You "can" build better reasoning AI Agents with Smaller models Let me explain how... MAD ( Multi-Agent Debate) is one of the few processes that utilize smaller language models to improve reasoning. Its process adapts a few core ideas from Meta's LLM as a judge and knowledge pool supporting MAD paper. However, this process is best if you have limited computational resources but still want a better reasoning system. Here's how it works: 1. User input: - User sends their query via text or any other multi-modal input. 2. LLM initial output: - The core LLMs evaluate the query and create a draft of the initial response. - The initial response is broken down into pieces so that SML can go through them even with lower context length. ?? Note: You want to make sure that this model has better reasoning capability (eg. Claude) and can call tools as well. ?? Note: You can use LRMs but a few LRMs like Deepseek R1 which is a cheaper option don't have a native tool-calling feature yet. Whereas, o1 and o3 do along with multi-tool coordination. 3. SLM-1 First response - The first small language model takes this input question and the answer generated by the LLM and tries to verify using its logic. - These are smaller models with fewer parameters which are only used to build upon the logic of the initial input. 4. SLM-2 Thinking and Response - SLM-2 utilizes a knowledge pool with tools to understand the context better. - When you're choosing SLM-2 you need to make sure that they have tool calling ability, the example shown GPT-4o mini supports tool calling. 5. Giving Improved responses - After utilizing tools, Knowledge Pool and Memory, the SLM-2 will critique if the output was wrong and give suggestions. 6. SLM-1 - SLM-1 will then reflect upon the new answers with the prompt given by the user and hence support or question the answer of SLM-2. - This process goes on "n" times until a conclusion is reached 7. LLM receives the verified response - The LLM finally receives the verified response allowing it to reflect on the initial response and build upon it. - Since the responses are in smaller segments, the LLMs build a comprehensive summary before sending out an improved repose to the user. Small Language Models like Phi-4 and Mistral Small 25.1 are suited for a wide range of tasks and are very cost-effective. Hence, this method builds upon the logic of using Smaller models to improve reasoning with collaborative contradiction. (Note: This is a simplified version, you can check out a deep coverage where Google uses the Sparse communication to improve the approach. Link in comments) What are your thoughts about this approach? Let me know in the comments below ?? Share ?? ? React ?? ? Save ?? & follow for everything related to AI Agents P.S Check comment for references

The Next-gen AI Agent stack is already here and you know it too Here's what the top companies are using to build AI Agents... When building an AI Agent, you can utilize all of your resources to build agents from scratch taking more time to build and scale. Or you can choose the route of what most businesses are using and use the most suitable frameworks to build it faster. These agentic stacks are very important if you want to build a sustainable agentic application for real-life use cases. Today, I am sharing the most used agentic stack which you too can use to build and ship faster. 1. GPU/CPU Providers - Responsible for computing resources as well as security - Popular providers include AWS, Azure, GCP, Groq, etc. 2. Infra/base - Responsible for the containerization and connectivity of your agentic application. - Popular providers include Docker, Kubernetes, Autoscale VMs, etc. 3. Database - Ensures agents store, retrieve, and manage data to maintain context and improve over time. - Popular providers include Chroma DB, Qdrant, Supabase, Pinecone, etc. 4. Foundation Models - Provides the brain of the operation, LLMs, LRMs and SLMs for different agentic use cases. - Popular providers include o1, o3, Deepseek R1, Gemini 2.0, Claude Sonnet 3.5 etc. 5. Agent Orchestration - Frameworks that lay the rules of how agentic interaction will take place, they manage the complexity of workflows and coordinate multi-agent interaction and much more - Popular frameworks include Langchain, CrewAI, Auotgen, AWS Multi-agent orchestrator, IBM Bee, Pydantic, etc. 6. Agentic Observability - Tools that help us keep track of agent observability, metrics, evaluations, and prompt management. - Popular providers include Langfuse, Helicone, Arize, Langsmith etc. 7. Tools - External services, APIs, and custom tools which allow agents to connect and build a network of services. - Popular Tools include Gmail, Duckduckgo search, Serper API, Exa, Composio, etc. 8. Memory - Builds the foundation of agentic short and long-term memory. - Popular Memory frameworks include: ZepAI, Mem0, Cognee, Letta etc. 9. Frontend - Frameworks that help in building a proper interactable UI for your Agentic application. - Popular frameworks include: Streamlit, React, Gradio, Nodejs, NextJs, etc. (Note: The list is not exhaustive, depending upon your use case - you need a set number of tools and frameworks) Which is your favourite tech stack to build AI Agents? Let me know in the comments below ?? Share ?? ? React ?? ? Save ?? & follow for everything related to AI Agents

From OpenAI Operator to Perplexity AI Assistants, Here's everything that happened in AI Agents last week ?? OpenAI released their very first agents Operator - https://lnkd.in/gQdqM6FN ?? Perplexity releases Perplexity assistant for all mobile users - https://lnkd.in/gQ8tHanu ?? Google releases in-depth research about a Data-Centric Framework ?for Self-Adaptive Agents - https://lnkd.in/gntBK7qh ?? Anthropic releases their Computer use course in collaboration with Deeplearning AI - https://lnkd.in/g5E4eDmg ?? Voiceflow release integration support for Voice AI Agents - https://lnkd.in/gzgrqp_m ?? Pinecone releases Pinecone assistant which will help you to AI Agents and RAG systems faster - https://lnkd.in/g6jUWf8V ?? GitHub Copilot Workspace is now available for everyone - https://lnkd.in/g5ivCH3q ?? Postman release their very first AI Agent builders - https://lnkd.in/gPwPSz-s ?? LlamaIndex introduces Agent workflow, a system to create AI Agents - https://lnkd.in/gmmVanTN ?? ByteDance releases UI-Tars, GUI Navigation using AI Agents? - https://lnkd.in/gEKuprEX ?? World Economic Forum released an article sharing why Manufacturers should adopt AI Agents - https://lnkd.in/gJmJfW-a ?? Devin releases their new update with personal browser integration - https://lnkd.in/g8vh2utU ?? Oracle shares a press release on how they improved customer connections using sales agents - https://lnkd.in/gEaBsrZh ?? A new full-stack app generator is released by Xiaoyin Qu, Heyboss - https://www.heyboss.xyz/ ?? Pig Dev is released, a browser for your AI Agents - https://www.pig.dev/ ?? Lovable collaborates with Builder Io to bring Figma designs into full-stack application? - https://lnkd.in/gs3SAN5Q ?? Check out last week's coverage: https://lnkd.in/gkWqPryk Which is your favorite update so far? Let me know in the comments below ?? ?? Share? ?? React? ?? Save & follow for everything related to AI Agents

What an amazing opportunity to kick off 2025 with my first conference of the year by speaking about Evaluating AI Performance Metrics at the 4th Annual AI in Medical Affairs conference https://lnkd.in/dyesGKyE in Philadelphia. Connect with me on LinkedIn if you are interested in customizing a 1-hour course for your teams on a deep-dive to evaluate, compare, fine-tune, and evaluate the bottom-line performance of the constantly evolving AI models in life sciences which will guide critical medical decisions. Honored to meet the speakers, leaders, peers, and innovators in our field - thank you for the thought-provoking discussions. #aimedicalaffairs #aihealthcare #aimetrics

We are 2 hours away from a Live Expert Panel Discussion about how to illuminate the purpose of measuring performance in Medical teams across pharma and biotech and continuously raise new standards in quality by setting new records in a 2nd of our Monthly Webinar Series at Amedea Pharma. Register here now https://lnkd.in/eT3VpAGK for access to slides you will be able to download after the session. The guest speakers Jeff New Luz Andrea Ramirez Lizbhet Delgado, PharmD Alex Pasek, MD have decades of experience leading Medical Teams, Operations, Budgets and Resources as well as Healthcare Change Management, Enterprise Software Company leadership and Strategic Partnership management at major global philanthropic organizations so join us live to ask them questions and learn from their case studies and insights. #highperformanceteams #medicalaffairs #performancekpis