Sundararaman Sankaran

A lot of machine learning work gets stuck in notebooks — but real value comes when we can share, test, gather feedback, iterate, and deploy. In this article, I walk through a Rapid ML Testing workflow: ✅ Build and test your model locally ✅ Add an interactive UI with Gradio ✅ Share instantly with Ngrok ✅ Deploy to Google Kubernetes Engine (GKE) ✅ Fully automated CI/CD for repeatable deployments This workflow helped me close the gap between experiment → working demo → production-ready service, while enabling fast stakeholder feedback. #MachineLearning #MLOps #Kubernetes https://lnkd.in/eeZdeT3Z

3

Excited to share Veeva Environmental Monitoring, a new application connected to Veeva LIMS and Veeva Quality Cloud designed to help labs and manufacturing facilities schedule, collect, and analyze environmental samples while staying aligned with rigorous sterility procedures and GMP standards. Many organizations still rely on disconnected, site-specific systems and manual processes for environmental monitoring. By integrating enterprise EM into LIMS and broader quality applications, teams gain visibility across sites and workflows, link results to quality events and procedures, and streamline batch release - all from a single cloud platform. #EnvironmentalMonitoring #QualityControl #QC #Vault

96

1 Comment

The SCDM India25 opening keynote in Hyderabad is framed around the question “Can agents automate your trial data workflow?” Key note speaker Anand’s “LLM psychologist” lens to examine how AI agents can augment or automate parts of clinical data management #scdmindia #ai #keynotespeech

103

GenAI & LLMs: What Every Business Leader Should Know: 🧠 LLMs generate text one token at a time; stopping is controlled externally. ✍️ Corrections don’t instantly update models — retraining is required. 🗂️ Memory features simulate recall via stored prompts, not real-time memory. 🌐 post-training knowledge gaps are filled via live data or retrieval systems. 📄 Uploaded documents can’t fully override training data. 📚 Citations may be fabricated — always verify sources. 🧾 Long context windows help, but RAG still boosts relevance and efficiency. 🧩 Memory personalizes responses but doesn’t fix factual errors. 🧠 LLMs focus on the beginning and end of prompts — middle content may be missed. ❌ Hallucinations can’t be eliminated but can be reduced with smart techniques. ✅ Efficient answer-checking blends automation (AI judges, unit tests) with human oversight. 🔁 Exact answer consistency isn’t guaranteed — caching helps, but variation remains. 📌 Strategic GenAI use demands understanding, not just results. More details in below article : https://lnkd.in/dChZ7jUD

2

A huge milestone for the Auxilius team and our customers.   When we started Auxilius, the vision was simple but ambitious: give biopharma finance and accounting teams the data foundation they need to manage their largest and most variable cost with confidence. Today, over 100 biopharma companies, from emerging biotech to big pharma, rely on Auxilius to do exactly that.   Joining the Veeva Product Partner Program is a real step forward in that vision, bringing clinical and financial data together so sponsors get more visibility into their R&D programs than they've ever had before. And it's a milestone we couldn't have reached without the customers who pushed us to keep building.   So grateful to our team (shout out Kristen, Peter, and Zach) and to the partners at Veeva who believed in this with us. Excited for what comes next.   More from Adam below.

39

Monday Fact: confusion is the only outcome when everyone is using AI to respond to one another. Case in point: RFPs. The bain of agency life - but also the lifeblood of new business. As you might imagine, they are great fodder for usage of AI. Need to write a 12 page RFP request document (client side)? Use your in house LLM! Need to respond to said 12 page document? Have an opportunity to ask Q+A? Yeah, AI (you now see request limitations in RFPs on the volume of questions that can be asked). Need to respond to the 120 questions you just got? You get the idea. Now, that’s where the fun really begins, because as you know, AI is really good at writing in generalities and vaugeries and giving the perception of competent questions and responses. But what you really see happen is that questions filled with vaguely reasonable content are responded to with answers of equivalent vagueness. In fact, they kind of compound each other. So by the time you, as the agency responsible for handing in the RFP, have to make sense of it, it’s very hard to make sense of. Things will likely continue to move ahead in this manner but the days of an RFP being a meaningful representation of your capability (on paper) are numbered. You’d better hope you get to the presentation stage so you can have a conversation with a real human!

11

2 Comments

Few weeks back, I asked where you have watched a data integration initiative fall apart. 835 people read it. 28 liked it. Nobody answered. So I will go first. I watched one fall apart at a company I worked at in the past, not because the platform failed. The platform was fine. It fell apart about eight months in, when the domain teams who were supposed to own their data products quietly stopped showing up to the working group. Nobody quit. Nobody said no. They just had an actual job to do, and this wasn't it. That is the thing nobody says in the post-mortem. The initiative did not fail. It just lost to everything else on the priority list. Every single week, for eight months, until there was nothing left to save. So here is where my thinking has landed. Every data architecture we have built in the last 30 years was designed for how we wished humans would behave. ·      Data mesh assumes people will voluntarily become data product owners. ·      Data catalogs assume someone will keep the metadata current. ·      Governance frameworks assume teams will consistently prioritize data quality over shipping. These are not unreasonable assumptions. They are just ones that rarely survive contact with a real world. So, I have been asking a different question. What if the execution model itself is the problem? What if we stopped designing systems that require sustained human behavior change at scale, and instead let agents handle the parts where human consistency has always broken down? That is the idea I have been developing. I am calling it Agentic Data Mesh. Not a new data architecture. A different assignment of work. ·      Agents discover the data where it lives. ·      Agents infer the domain it belongs to. ·      Agents draft the data product definition. ·      Agents enforce governance and document the audit trail. Humans provide judgment at the decision points that actually need it. They do not have to build from scratch or maintain things continuously. That changes the ask from a sustained organizational commitment to a series of discrete approvals. In regulated industries like Pharma, this matters even more. Every connection, every transformation, every decision, documented, traceable, inspection-ready. Not because someone remembered to write it down. Because the process itself generates the record. I think the data silo problem was never really a technology problem. It was always an incentive and ownership problem. Agentic Data Mesh is my attempt to propose an architecture that works with how organizations actually behave, rather than assuming they will start behaving differently once the right framework is in place. I am going deep on this. More to share soon. If this resonates, or if you think there's a flaw in the premise, I would genuinely like to hear it. #AgenticDataMesh #DataMesh #PharmaAI #LifeSciences #DataStrategy #CDO #EnterpriseAI #AIAgents #AgenticAI #Data #DataAI

35

12 Comments

Human + AI + Impact (Post-Conference Reflection) Failure is Data. Hype is Noise. Progress is a Choice. After our session at the Pharma CI Conference USA 2025, I’ve been reflecting on one key truth: AI isn’t replacing human intelligence — it’s amplifying it. Together with my colleagues Evgeny Efuni and Linh Nguyen from C5i, we explored how Human + AI collaboration is transforming Competitive Intelligence — turning fragmented signals into timely, actionable strategies. We referenced the MIT State of AI 2025 report — not as a warning, but as a mirror. The takeaway wasn’t about the “95% of GenAI pilots failing” — it was about what that number reveals: that success in AI isn’t about technology maturity, but about human alignment, feedback, and leadership. As Anees Merchant aptly said, failure isn’t a crisis — it’s innovation in progress. The real challenge isn’t AI itself, but how we learn, adapt, and lead with it. In CI and beyond, the winners won’t be those who avoid failure — but those who learn fastest and turn insight into impact. #PharmaCI #ArtificialIntelligence #HumanAI #Leadership #Innovation #LearningCulture #DigitalTransformation #CompetitiveIntelligence #CIPharma2025 #C5i #HumanAIImpact Ashwin Mittal Nitesh Jain Chandrasekhar Sharma Suchitra Eswaran Anees Merchant

23

🔔"There Is No A.I. - A Scientific Exploration" is now available on Kindle worldwide. Reason, intelligence, logic, and knowledge are not synonymous. - Howard Gardner. Fantasy: `AGI- Artificial General Intelligence`. A hypothetical entity which generalizes a certain behavioral aspect which is favorable to the creator of the entity. Reality: Howard Gardner, a renowned psychologist, pressed forward the idea of multiple intelligences in his seminal book “Frames of Mind”. He categorically calls out the idea of a single unified view of intelligence in humans (often equated with IQ) as overly narrow, empirically limited, and biased. The tech giants promoting this idea of an Artificial General Intelligence, which can overshadow the capacities of human intelligence, seem sinister as it celebrates certain traits, such as logical reasoning or linguistic capabilities based on probability not creativity, and at the same time downgrades other human abilities, such as musical, spatial, or bodily-kinesthetic capacities. Remember, the models can not conjure any magic but will produce output only based on the data they have been trained upon and the parameters (which in vague terms can be equated to narratives) are set by the model creators. In short, the outcome presented will be governed by the worldview of the model creator, and if humans start submitting their agency to AI models, it will eventually end up in a world with narrow view point which would be mostly manipulated at the whims of players behind the scene of the tech. History: This view of adhering to a general intelligence system is not new and has already led to educational systems that prioritize a narrow set of skills, disadvantaging students with strengths in other areas. These are nothing but signs of a society governed by authoritarian leaders who aim to radically transform their societies according to a specific ideological or political agenda. Their underlying aim is to seek total control over societal structures, often suppressing dissent to enforce a unified vision. And it has often been done historically by alienating entities associated with reason, critical thinking, and intellectual inquiry. Evidence: A 2020 New York Times article notes that lawmakers accused Google, Amazon, Facebook, and Apple of serving as “gatekeepers” that make third-party businesses beholden to their demands, with the term “monopoly” appearing nearly 120 times in a congressional report. Quest to monopoly is not a bad thing, as Peter Thiel suggests, it is the condition of every successful business. In an ideal yet dynamic world, creative monopolists give customers more choices by adding entirely new categories of abundance to the world. But the line between ‘choice’ and ‘control’ is getting blurred day by day in the new ecosystem that is at play. Rediscover Intelligence.

13

1 Comment

In the rush to get to market, commercial strategies often break at the point of handoff. Speed is rewarded — faster decisions, faster launches, faster field readiness. But when timelines compress, change management is usually the first thing sacrificed, not because it’s unimportant, but because it’s invisible on a Gantt chart. The result is a strong strategy entering an organization that isn’t ready to absorb it. Roles blur, assumptions go untested, and field teams are asked to move quickly without the clarity or infrastructure to do so. What follows looks like an execution problem: inconsistent adoption, delays, uneven performance. In reality, it’s a design issue created upstream in the name of speed. The commercial organizations that scale well don’t choose between speed and adoption. They design for both — building operating models that can move fast and land change effectively in the field. This tension shows up across industries wherever scale, speed, and complexity collide. #commercialstrategy #changemanagement

57

1 Comment

AI will only transform clinical research when the data is RIGHT. Everyone in life sciences is racing to implement AI. Many might be disappointed. Not because the technology isn't ready. It is! But because you cannot build intelligent systems on top of fragmented, ungoverned data — and most organizations haven't done the unglamorous work of fixing that first. I've seen this pattern play out across 20+ years leading enterprise technology programs. The companies that extract real value from AI share one thing: they treated data as a strategic asset before they treated AI as a solution. Before deploying ML and NLP across regulatory and clinical workflows, we had to architect a governed data lake spanning 250+ systems. That foundation — not the algorithms — is what drove 35–55% efficiency gains. The AI roadmap and the data governance roadmap are the same roadmap. Three things I'd tell any life sciences technology leader right now: → Your AI strategy is only as strong as your data foundation → Governance isn't a compliance exercise — it's a competitive advantage → The CRO or pharma company that wins the next decade will be data-native first, AI-enabled second The technology is not the hard part. The discipline to build it right is. #AIinHealthcare #LifeSciencesIT #DigitalTransformation #ClinicalResearch #DataGovernance #CRO #Pharma

265

21 Comments

𝗔 𝗥𝗲𝗮𝗹 𝗙𝗗𝗔 𝗖𝗮𝘀𝗲 𝗧𝗵𝗮𝘁 𝗦𝗵𝗼𝗰𝗸𝗲𝗱 𝘁𝗵𝗲 𝗣𝗵𝗮𝗿𝗺𝗮 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 Sharing an insightful video by Manne Venkanna Chowdary, CEO of Amplelogic, highlighting how improper use of AI can lead to serious compliance risks in the pharma industry. 𝗪𝗮𝘁𝗰𝗵 𝘁𝗵𝗶𝘀 𝘁𝗼 𝘀𝗲𝗲 𝘄𝗵𝗮𝘁 𝗿𝗲𝗮𝗹𝗹𝘆 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱 : https://lnkd.in/gs9vVFHN As AI adoption accelerates, ensuring data integrity, validation, and regulatory compliance becomes more critical than ever. This case is a strong reminder of the responsibilities that come with innovation. 𝗪𝗼𝘂𝗹𝗱 𝗹𝗼𝘃𝗲 𝘁𝗼 𝗵𝗲𝗮𝗿 𝘆𝗼𝘂𝗿 𝘁𝗵𝗼𝘂𝗴𝗵𝘁𝘀 𝗼𝗻 𝗵𝗼𝘄 𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻𝘀 𝗰𝗮𝗻 𝗯𝗮𝗹𝗮𝗻𝗰𝗲 𝗔𝗜 𝗮𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗰𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲. #Pharma #FDA #AI #Compliance #DataIntegrity #LifeSciences #Amplelogic #PharmaceuticalIndustry #PharmaCompliance #RegulatoryCompliance #USFDA #FDACompliance #LifeSciencesIndustry #GxP #GMP #DataIntegrity #QualityAssurance #QualityManagement #QMS #DigitalTransformation #PharmaTech #AIInPharma #AICompliance #ResponsibleAI #Validation #CSV #AuditReady #InspectionReady #RiskManagement #Pharmacovigilance #ClinicalResearch #MedTech #Biotech #HealthTech #AutomationInPharma #ComplianceMatters #InnovationInPharma

5

🔍 Why Consistency Matters in Research — A 30-Year Perspective from EMA Enterprise Management Associates just marked a major milestone — 30 years of independent, practitioner-centric research. The blog, “EMA at 30: The Value of Consistency,” cuts straight to what sets trusted research apart in today’s crowded information landscape: clarity, context, and credibility. In an age where buyers are overwhelmed with reports, webinars, demos, and AI-generated content, consistency isn’t just an operational virtue — it’s a strategic differentiator. EMA’s decades-long commitment to: • independence over scale • technical depth over superficial breadth • human judgment over algorithmic noise undergirds research that helps buyers and technology leaders make sense of complexity rather than just adding to it. As markets confront AI disruption, cloud transformation, security threats, and regulatory change, consistent, experience-anchored insight is no longer a “nice-to-have” — it’s required for decision confidence. 🧠 Insightful reflection and well-anchored perspective — worth reading for any enterprise technology leader. Read the full blog here: 👉 https://lnkd.in/gUnEzHzC #ema #research #technologytrends #ITleadership #enterpriseIT

22

3 Comments

I wanted to share this profound insight from Datavant’s Arnaub Chatterjee on the latest episode of Medidata Solutions' podcast "from Dreamers to Disruptors" with our CEO Anthony Costello. There's a critical opportunity at the intersection of sites and trial design to reduce burden and reshape patient recruitment processes. I couldn’t be prouder of our Site Insights Program, where life sciences visionaries come together to build a superior site experience for everyone. Check out the full conversation here: https://mdso.io/g2r #Medidata

53

🚀 Exciting milestone in Life Sciences technology — Veeva Vault Quality AI Agents are live this April 2026! As someone deeply embedded in Veeva Vault implementations at Rudhra Info Solutions, I'm thrilled about the April 2026 rollout of Veeva Systems AI Agents for Quality and Safety — a game-changer for life sciences quality management. 🔍 What are Veeva Quality AI Agents? Unlike traditional automation, AI Agents act as proactive collaborators — they understand your Quality System (SOPs, CAPAs, Change Controls, Deviations) and take initiative, not just react. ⚙️ Key capabilities rolling out: ✅ Trend detection in Deviations & CAPAs — flags issues before they escalate ✅ Root cause suggestion based on historical batch records ✅ Automated audit & inspection readiness summaries ✅ SOP change validation against regulatory updates ✅ Supplier and batch data monitoring for emerging risks ✅ Works within established user access controls, permissions & audit trails 🏗️ Architecture powering it: Built on Amazon Bedrock (Anthropic + Amazon LLMs), with support for Azure AI and customer-provided models — fully configurable within @Veeva Systems Vault Platform. 📅 2026 Rollout Timeline: ✅ December 2025 → Commercial (Vault CRM, PromoMats) 🔥 April 2026 → Safety & Quality ← We are HERE August 2026 → Clinical Operations, Regulatory, Medical December 2026 → Clinical Data 🏢 What this means for Quality & Compliance teams: Proactive risk management backed by AI-driven data insights Faster audit readiness with intelligent documentation support Compliance with 21 CFR Part 11 and EU Annex 11 built into the workflow Reduced manual oversight burden on QA/Regulatory teams 💡 At Rudhra Info Solutions, we are already preparing our Veeva Vault environments to leverage these AI agent capabilities for our life sciences clients. This isn't just a feature update — it's a fundamental shift in how QA operates. AI agents embedded in your core quality workflows means speed + oversight, not a trade-off between them. Are you preparing your QMS for the Veeva AI Agent rollout? Let's connect and share experiences! 👇 #VeevaVault #VeevaAI #QualityManagement #LifeSciences #AIAgents #Pharma #GxP #CAPA #Compliance #DigitalTransformation #QMS #MedTech #HealthcareAI #RudhraInfoSolutions

37

🧪 Building a Reliable Chemical Reaction Data Foundation for Modern Drug Discovery 🔬💡 In today’s accelerated R&D landscape, the quality and reliability of chemical reaction data have become fundamental to innovation in Pharma & Life Sciences. From AI-driven molecule generation to predictive synthesis planning, everything begins with one core element—trustworthy reaction datasets. A robust reaction data ecosystem is no longer optional—it’s a strategic enabler for next-generation drug discovery and competitive advantage. 🚀 🔗 Read the full article here: https://lnkd.in/ggDdAMuF How prepared are we as an industry to elevate reaction data quality for the AI-driven future? #DrugDiscovery #ChemicalReactions #DataQuality #PharmaR&D #LifeSciences #ComputationalChemistry #AIinPharma #DigitalScience

1

Biologics are changing everything about how medicines are packaged, stored, and moved. This piece on “The Future of Biologistics” explores how the rise of mAbs, mRNA, cell and gene therapies, and other advanced modalities is forcing pharma to rethink cold chain, real-time monitoring, and end-to-end visibility across the supply chain. At USDM Life Sciences, we help organizations build the digital and compliance foundation needed for this next era—connecting quality, labeling, serialization, and GxP cloud platforms so biologics can move safely, compliantly, and at scale.

2

RJB Consulting Partners LLC and FocalCXM are proud to announce a strategic partnership focused on accelerating commercial innovation across the life sciences ecosystem. By combining executive-level commercial strategy with next-generation AI, data, workflow automation, and customer engagement capabilities, we help biopharma organizations modernize operations, strengthen customer experiences, and drive measurable business impact. The mission is simple: deliver faster, smarter, and more scalable commercial solutions that create real value for clients, patients, and partners alike. For partnership inquiries, executive presentations, or commercial strategy discussions, please contact Robert.Britting@FocalCXM.com RJB Consulting Partners LLC Robert Britting Jr. PhD, MBA Adithya Buddhavarapu FocalCXM

10

2 Comments