Go Summarize

No Priors Ep. 25 | With Palantir's CTO Shyam Sankar

7K views|9 months ago
💫 Short Summary

Sham Sankar, Palantir's CTO, shares his journey from Nigeria to Palantir, discussing the company's platforms and products like AIP, Gotham, and Foundry. Apollo, LMS, and AI tools are highlighted for data integration and decision-making. The importance of building customer experiences, managing resources, and incorporating diverse viewpoints is emphasized. Future UI integration trends towards minimal UI and language-based interactions. The significance of user customization, system integration, and trust in complex systems is discussed. Palantir's focus on Healthcare and optimizing workflows through AI for alerts and state machines is also highlighted.

✨ Highlights
📊 Transcript
Sham Sankar discusses his background and journey to becoming Palantir's CTO.
He grew up in Nigeria, moved to the US as a refugee, and started as the 13th employee at Palantir after working at Zoom.
Sankar is excited about Palantir's new platform AIP and the opportunities it presents.
His involvement with Palantir came through connections in Silicon Valley and seed fund investors.
Palantir is known for its secretive nature and initially had a small community in technology circles.
Journey from Small Company to Forward Deployed Engineering.
The speaker started at a small company related to terrorism, eventually becoming employee number 13 and taking on multiple roles.
Forward deployed engineering involves computer scientists interacting with customers to create a hybrid role of product management, customer success, and engineering.
The focus is on solving customer problems and making a positive impact, rather than just meeting technical specifications or past ambitions.
Overview of Palantir's government products.
Gotham is designed for intelligence and defense customers to integrate and model data for decision-making.
Foundry is a data integration platform that transforms structured and unstructured data into code and creates a semantic layer.
The semantic layer models both nouns and verbs, providing a digital twin for conceptual understanding in decision-making.
AIP is part of the ecosystem, with Gotham at the top and Foundry as a foundational platform for data integration and modeling.
The platform integrates with SAP, allocates inventory, models enterprise decisions, orchestrates transactional systems, and provides simulation capabilities.
It was utilized for COVID vaccine distribution and supply chain management during crises.
Initially focused on intelligence and defense clients before expanding to commercial clients.
Engineers were inspired by projects like James Bond to develop software for integrated data.
The Apollo platform was created for air gap environments, deploying modern software without internet access.
Apollo is a successor to cicd, focusing on autonomous software delivery and deployment.
It allows separate modeling of software and environments, managing upgrades and dependencies.
The approach includes blue-green upgrade patterns, health checks, vulnerability management, and software recalls.
Apollo is gaining traction in complex environments like air-gapped customer environments and European Sovereign SAS deployments.
AIP is a core technology enabling AI-powered experiences within private networks, driving decision-making and AI-enabled applications.
Importance of investing in Learning Management Systems (LMS) with a focus on ontology for maximizing value.
Dynamic ontologies and semantic layers are essential for compressing information and creating reliable functions.
Engineers encounter challenges with LMS similar to statistics and calculus, necessitating a comprehensive toolchain for managing stochastic elements.
The ontology grounds LMS in reality and business context, simplifying management processes without changing the model.
Importance of Execution and Planning in Project Building.
Emphasis on calibration and selecting the right use cases for telemetry and production log data.
Leveraging existing ontologies for handling messy data effectively.
Example shared on generating courses of action from operational plans in defense.
Mention of leveraging AI tools and ontologies in industries like Pharmaceuticals and defense.
Importance of managing resources, risks, and assumptions in application development.
Common operating pictures are essential tools for effective management.
Limitations of chat interfaces in the application development process.
Building customer experiences is crucial, especially in commercial applications like quality and warranty claims for auto manufacturers.
Models play a significant role in driving experiences and building trust in various applications.
Utilizing 'mad geniuses' to solve complex problems.
Emphasizing the importance of diverse viewpoints in problem-solving.
Exploring the potential of overlaying assistance on existing software and creating automatic workflows.
Visualizing the combination of application states and user intent to generate new application states.
Highlighting the complexity and artistry involved in the process.
The future of UI integration may involve minimal UI with agents representing users, interacting with APIs programmatically.
Using language instead of UI can drastically reduce development time, as seen in a case where a feature took hours instead of months to implement.
This shift in thinking allows for the creation of integrated single-pane experiences using backend tools and primitives.
It changes the way applications are built and interacted with.
Benefits of allowing users to customize UIs for better user experience and efficiency in the Enterprise context.
Challenges of generalizing solutions for different customer needs and implementing specialized data or integrations can be time-consuming.
Importance of simplifying and customizing systems to efficiently meet specific needs.
Transformative nature of stringing APIs together for improved functionality and ease of use.
Significance of data rights and the ability to fine-tune production designs for better management and maintenance.
The importance of integrating various components in a complex system and developing trust in the system.
Emphasis on the need for a robust evaluation environment and making co-pilot models accessible to everyday users.
Addressing technical challenges to ensure smooth workflow and user trust in the system.
Significance of system integration and ongoing work to improve usability and reliability.
Importance of Defensibility in Non-Traditional Code Approaches
Challenges in Customer Education regarding new concepts and validating outputs.
Understanding the state machine and the role of agents in the enterprise is crucial.
Agents have authority over specific state transitions.
Transition from individual contributors to managers of agents is emphasized.
Importance of trust in log data for change management and comparison to Tesla's self-driving approach.
Emphasis on incremental self-driving improvements and feedback for success.
Palantir's strategy of onboarding people with new technology through hackathons and experimentation.
Organizational focus on experimentation and integration of new tools for improved productivity.
Importance of Ambition and Aggressiveness in Pursuing Goals.
Emphasizes the need for constant innovation and adaptation in a world with unpredictable solutions.
Value of using multiple models simultaneously to understand outputs and explore capabilities.
Applying emerging technologies to solve crucial business problems is essential.
Decision-making within a company is interconnected and must be approached with this in mind.
The potential of using films and generative AI in healthcare for enhancing patient care and operational efficiency.
Palantir has dedicated a third of their business to the healthcare sector, working with global customers like the NHS in the UK and multiple hospital systems in the US.
AI in healthcare is considered a major growth area, impacting clinical care and operational processes.
The challenges and adoption pace of AI in healthcare may differ across various use cases in the industry.
Prioritizing alerts and turning them into recommendations and staged scenarios for evaluation.
Moving from surfacing alerts to providing solutions, reducing human toil, and improving operational workflows.
Optimizing state machine throughput and claims processing workflows.
Creating analytical workflows for managers to manipulate data structurally, rather than expecting AI to change state machines directly.
The importance of thresholding and learning in processing alerts.
LLm can help prioritize actions based on consequences rather than severity.
This approach improves efficiency and strategic response to alerts.
Positive overview of discussed topics and initiatives.
Gratitude expressed for insightful discussion.