What needs improvement with Palantir Foundry?

I think the things that I do not like about Palantir Foundry is not a Palantir issue so much as it is from my company side and what they have commissioned for and have not commissioned for. With Palantir Foundry, it is a part of the user tools that they provide. Their AI Assist that they use is something I have found that sometimes I get better results for when I do need help and aid. I get better results going to outsourced AI assistance sites such as Gemini, because it seems like AI Assist, which my understanding is it is supposed to be searching and utilizing the documentation for Palantir Foundry, but sometimes it gets kind of confused in the capabilities and it will tell you, 'Oh, you can use this and do this.' And then when you try it, the system says, 'We do not support this.' Whereas I can go to Gemini and it will give me a workaround that will actually work.

Palantir Foundry needs two points for improvement regarding the data product. First, Palantir Foundry needs to improve their clear resume about their product features roadmap. Second, Palantir Foundry needs to have a closer connection with the enterprise corporate application, which means the business application, because big companies have a very huge ecosystem of business applications. In my personal perspective, I think Palantir Foundry still has some space to improve in integrating with the IT landscape of the corporate. I want to say that Palantir Foundry is quite expensive. It is not so easy for budget review and budget transformation of the company, which is quite expensive.

Palantir Foundry could be improved by addressing the need for some coding in the Workshop since we cannot expect 100% no-code functionality, especially when dealing with dynamic user input, which requires writing functions in the Code Repository.

The Workshop application could be improved because it is not very customizable, but it is still very strong. We also have the React OSK apps, but it does not allow the inbuilt applications such as Maps and AIP Analyst and everything else inside the React OSK. They should make Workshop a little more customizable or should add the applications to be embedded into the React app.

Palantir Foundry can be improved if cross-cloud integrations could be enhanced, as it would be very helpful and efficient. Currently, it has some limitations. From my experience in my last organization, there were some limitations where we could not connect directly to different cloud storage services, and we could not use those storage services for our backing data sets for the ontology. This area can be improved. There is a learning curve to Palantir Foundry, and better documentation would really help because while documentation exists, not everything is covered. Better training programs would really help new learners adapt to and grasp Palantir Foundry concepts.

Regarding how Palantir Foundry can be improved, it is evolving by adopting different other technologies, but one of the key factors is that it all depends upon the vision of the product. However, Palantir Foundry's license fees comparative to others are quite high. Secondly, Palantir Foundry comes with a different set of products. Apart from Foundry, there is Apollo and Gotham. I can understand the tight coupling between Gotham and Apollo because they cater to more critical missions. But from a perspective of Foundry, it can be a bit loose in terms of adopting different other technologies so that the use case can be widened. Integration is where they can improve. Support is fine, but the integration to different ecosystems which might be legacy is one area for improvement. One of the major points is pricing.

I see that Palantir Foundry can be improved in terms of tools that are not very customizable. For instance, the Workshop application has some widgets inside that we wanted to modify, but when we reached out to Palantir, they informed us that it was a design decision and those modifications would not be added. Therefore, the customization features in some cases are not accessible to end users, and it is up to Palantir to decide whether such modifications would be introduced or not. Palantir Foundry is quickly catching up on onboarding a lot of AI tools, which will be beneficial to the organization in the long run.

One challenge regarding how Palantir Foundry can be improved is the learning curve. Foundry has a very broad ecosystem with Ontology, Pipeline Builder, Code Repositories, and AI integrations. For new engineers or business users onboarding, it can take time, especially if they are coming from more traditional data platforms. Better documentation, simplified onboarding paths, and more beginner-friendly examples would help accelerate adoption. Another area is debugging complexity. While lineage and monitoring are strong features, troubleshooting deeply interconnected pipelines can still become difficult in a large enterprise environment. Sometimes error logs and pipeline failure messages could be more descriptive or developer-friendly, especially for distributed PySpark jobs. Another pain point is customization limitations in certain UI-driven components. While low-code tools are great for rapid development, highly customized workflows sometimes still require engineering workarounds or deeper technical implementation. The platform is extremely capable, but improvements around usability, debugging experience, DevOps flexibility, and ecosystem openness would make it even more effective for enterprise engineering teams.

I don't see any feature, process or aspect that I would like to see optimized or changed in Palantir Foundry at the moment.

When we were using ETL with Palantir Foundry, we found we had less freedom compared to Cloudera, where we had more liberty in using various configuration parameters of Spark, allowing us to tune our jobs accordingly. In Palantir Foundry, we realized that ETL is not a great use case while creating agents and dashboards was better. We faced many challenges in ETL that we did not face earlier in a big data Hadoop environment, such as Cloudera; we eventually removed that ETL process. I feel that if developers had more freedom to experiment, it would enhance Palantir Foundry. Specifically, if the ETL process could provide more freedom concerning Spark configuration parameters, that would be beneficial. I worked on three parameters: dashboards, ETL, and creating agents. The latter two functions really work well, but we faced challenges specifically in ETL. Additionally, the costs are substantial; I believe making Palantir Foundry more approachable for the general public would be beneficial.

There are some limitations when dealing with a large-scale project. It might not be very feasible to do so because we want to optimize it to a very micro level. There are some limitations to that when using Palantir Foundry.

I wouldn't add more about the needed improvements, either on the technical side or regarding compatibility and integration.Obviously the company's reputation needs to be improved regarding Palantir Foundry, or ideally, Palantir needs to get away from the appalling views on human rights and improve the reputation. Whether it can be improved, I don't know. This is not a technological problem; it is a problem of company image, so I wouldn't be surprised if the NHS actually triggers a break clause in the contract in February next year. That is not linked to the product itself. From a technical perspective, maybe to make Palantir Foundry more compatible with Databricks could be one option, or maybe more integrated with Azure. It is difficult to say because they might lose some of their competitive advantage in doing so.

I believe Palantir Foundry could improve by introducing a tool to restrict object-level creation to specific people, such as developers. A dedicated application could streamline requests for access to data across different organizational verticals, enabling better tracking of costs associated with specific use cases and improving identification of data access requests. Regarding documentation, I find that when I face issues, the outdated documentation is not helpful; for example, while trying to create a webhook to fetch SharePoint metadata, I found available resources lacking relevance, needing significant updates to assist users properly.

Regarding points for improvement for Palantir Foundry, I see that they are improving day by day. In the last one to two years, I have seen many improvements compared to the two years that I have worked on Palantir Foundry. There are many things that come up, but a few things are not intuitive enough. Now that we are in this AI phase, Palantir Foundry has created some wrappers around the models, allowing us to create using a no-code application, chatbots, and LLM functions. The problem is that interaction with outside applications can be difficult with the current setup that Palantir Foundry has. There are ways to do that, but it is not that intuitive, which is what I feel.

Apart from the pricing and offline availability issues, improvements are needed in Palantir Foundry's costing factor. Cost-wise, it is not open for everybody, and they are not exposing anything outside of the box. The major hindrance with Palantir Foundry is that being a very closed product, the cost optimization and costing are not exposed to the end users. Apart from that, it is a very good tool and product.

Palantir Foundry is missing marketing, which could help it grow. Additionally, the startup pricing is high, causing concern despite being cost-effective in terms of total cost of ownership. Palantir Foundry also needs to change the traditional data management approach from one-directional to bi-directional, near real-time data flow everywhere, which they address through data virtualization.

The solution’s data security could be improved. We cannot use many Python packages with the solution. We were able to use only a few compatible Python packages.

Palantir Foundry is very good for someone technical. The tool still needs to work on the non-technical part, where people can use its flexibility. The business user should not end up writing huge queries to get small snippets of data. The solution's visualization and analysis could be improved.

The solution's pricing is high. Compared to other hyperscalers, Palantir Foundry is complex and not so user-intuitive. There could possibly be a little bit of overhead concerning the maintainability of the platform.

Computing is very expensive. If you want to create new models on specific data sets, computing that is quite costly. Python's current setup within Palantir is very limiting. I would like to have more freedom to use Python without limitations.

The application development aspect can be improved significantly and would make a difference. We use some third-party tools for reporting and it's a challenge for us to move data into the file system because Palantir is a closed environment and there are difficulties receiving data from external sources. There are some options in place for dealing with that but it's not sufficiently intuitive. I'd like the data exporting functionality to be as intuitive as the importing.

The one area where improvement could be made is the cost of the solution which is quite expensive.

The data lineage was challenging. It's hard to track data from the sources as it moves through stages. Informatica EDC can easily capture and report it because it talks to the metadata. This is generated across those various staging points. It was hard to generalize that and put it into a catalog for people who you didn't know to reuse data. Maybe it's different now. That's the nice thing about Informatica: The catalog is reusable. Palantir is successful, and the institution loves them. I don't want to disparage it. I'm just speaking technically from an interoperability perspective.

The workflow could be improved. Although it works rather seamlessly, the workflow too complicated sometimes. Maybe they can reduce the complexity of the workflow. It could be more modularized in the future. The performance of the engine could be better.

They do not have a data center in Europe, and we have lots of personally identifiable information in our dataset that needs to be hosted by a third-party data center like Amazon or Microsoft Azure. There are some issues with scalability because when we are using a really large dataset, the system is rather slow. The performance can be improved. It would make our life a lot easier if it were as fast as Google Cloud. The GCP is unmatchable in terms of the speed at the moment. From a user perspective, it would be nice to have a preview of what the data is looking like. As it is now, you can see the schema but not the actual data. For example, they can see the different columns but they don't know what's there. If they could inspect the first few hundred columns of data then they would have an idea of what they are dealing with.