Hi folks I don't get the fastest and easiest way to turn a JSON into many columns without having to do it manually see LOOM: https://www.loom.com/share/933f5cf5b49f4723af2bb9d14895ca0c What I have done is extracted from a table to another the 10 lookalike from Ocean.io And then I'd like to turn all the JSON in this new table into all the right columns in 1 click without having to add each column individually Thanks
{ "id": "4bdf2e6a14d52f1e", "name": "Conrado M.", "state": "California", "domain": "cloudera.com", "origin": { "type": "list", "tableId": "t_0t3ccgvkndG4DYEdmgD", "recordId": "r_6jK52calIctx" }, "parent": { "Org": "Collibra", "Name": "Eric Warner, PhD", "Title": "Director, AI Engineering", "Country": "United States", "Summary": "", "Headline": "Director, AI Engineering", "LinkedIn URL": "https://www.linkedin.com/in/eric-warner-phd-6a249a47/", "Name - People": "Conrado M.", "Domain - People": "cloudera.com", "Country - People": "us", "Url - Experience": "https://www.linkedin.com/company/collibra", "Jobtitle - People": "Senior Director of AI Engineering", "Company - Experience": "Collibra", "Linkedinurl - People": "https://www.linkedin.com/in/conradomiranda", "Seniorities - People": "Director", "Companysize - Company - People": "1001-5000" }, "skills": [ "Programming", "Matlab", "Signal Processing", "C++", "Statistics", "Algorithms", "Python", "Artificial Intelligence", "Simulations", "C", "Computer Science", "Modeling", "Control Systems Design", "Embedded Systems", "LaTeX", "Optimization", "Evolutionary Computation" ], "company": { "logo": "https://cdn.ocean.io/logos/cloudera.com/883cd4fb5c8b94df345e6cc4a60c7e9a41ef256a.png", "name": "Cloudera", "companySize": "1001-5000" }, "country": "us", "summary": "At Verta, we're changing how companies do machine learning. Our end to end ML platform focuses on a streamlined experience from experimentation to production. Contact us if you want your company to take the lead on building and deploying models, the right way. We're hiring! If you're interested in joining a small and fast team focused on scale, robustness, and flexibility for our customers, let's talk.", "jobTitle": "Senior Director of AI Engineering", "lastName": "M.", "location": "San Francisco Bay Area", "firstName": "Conrado", "departments": [ "Management" ], "experiences": [ { "domain": "cloudera.com", "dateFrom": "2024-06", "jobTitle": "Senior Director of AI Engineering", "linkedinCompanyHandle": "cloudera" }, { "dateTo": "2024-06", "domain": "cloudera.com", "dateFrom": "2019-07", "jobTitle": "Chief Technology Officer", "description": "Acquired by Cloudera<br>At Verta, we provided a platform for enterprises to bring ML and AI to the real world, while minimizing their risk and TCO.<br>- Led product & infra for LLM evaluation, distillation, safety checks, and high-throughput inference pipelines.<br>- Built AI Evaluation Workbench enabling systematic comparison and benchmarking of multiple models and prompts through LLM-as-a-judge and human evaluation pipelines. Platform iteratively collaborated with users to understand<br>evaluation criteria, then auto-generated granular test queries and validation checks, cutting experiment cycles 10× and guiding Fortune 100 GenAI roll-outs..<br>- Built secure MLOps platform to deploy & monitor 1000+ models with continuous safety/compliance checks.<br>- Closed major enterprise deals using MEDDIC/Challenger technical sales strategy and trusted advisor role.<br>- Recruited and developed diverse, high-retention team from L8 to new grads with 100% offer acceptance. Successfully transitioned company from in-office to fully remote during COVID while maintaining team performance and cohesion.", "linkedinCompanyHandle": "vertaai" }, { "dateTo": "2019-07", "domain": "nvidia.com", "dateFrom": "2018-06", "jobTitle": "Engineering Manager", "description": "I was an engineering manager in the AV platform (MagLev), handling DL SDK, CI, and large scale inference. I worked across the 6+ organizations collaborating in AV to align roadmaps and priorities. My team focused on building developer tools for scalability, reproducibility, and DL in production systems, and in massive scale evaluation of the self-driving stack for detect failures and business KPIs.<br>My team:<br>- Moved all DL AV applications to a new cluster transparently, while providing 4x training speed-up.<br>- Unified data representation, processing and access for DL models.<br>- Developed the first workflow system for MagLev, enabling AV developers to orchestrate complex tasks.<br>- Achieved 10x CI speed-up, 20x PR merging speed-up and <.1% failure rate with Bazel, distributed cache and merge queues.<br>- Unified development environments between personal machines, CI and cloud, ensuring consistent behavior across the computing systems.<br>- Created and open-sourced a new library for Python management with Bazel, being quickly adopted by other companies.<br>- Developed and maintained the stack to enable NVIDIA's Drive to run in containers and cloud.<br>- Built analytics tools to allow AV teams to identify driving failures and guide model improvements.", "linkedinCompanyHandle": "nvidia" }, { "dateTo": "2018-05", "domain": "nvidia.com", "dateFrom": "2017-11", "jobTitle": "Staff Software Engineer", "linkedinCompanyHandle": "nvidia" }, { "dateTo": "2017-11", "domain": "x.com", "dateFrom": "2016-07", "jobTitle": "Sr. Software Engineer", "description": "- Tech lead for the deep learning platform used across Twitter (DeepBird) by 10+ teams.<br>- Architect and lead for the new platform design, transitioning from Lua-Torch to Python-TensorFlow for ease of use and flexibility.<br>- Co-authored framework to unify data science and machine learning at Twitter. It facilitates development of Scala and Python code with notebooks and faster deploy, job configuration on the cluster and scaling with Spark and Scalding.<br>- Distributed word and user embedding pre-processing with Spark, resulting in >10x speedup, flexibility to scale to >2000 cores and support for >3x vocabulary sizes.<br>- Optimized GBDT training for dense datasets, achieving 2-5x speedup over XGBoost.<br>- Reduced CPU consumption and latency by 25% for models used by Twitter’s home timeline without change in predictions quality.", "linkedinCompanyHandle": "twitter" }, { "dateTo": "2011-07", "domain": "encora.com", "dateFrom": "2011-01", "jobTitle": "Developer", "description": "Development and maintenance of TDM to VoIP gateways.", "linkedinCompanyHandle": "daitan-group" }, { "dateTo": "2014-12", "domain": "unicamp.br", "dateFrom": "2010-08", "jobTitle": "Teaching Assistant", "description": "Teaching Assistant of<br>- Embedded Systems Design;<br>- Mechanical Systems Control;<br>- Signal Processing;<br>- System's Advanced Control.", "linkedinCompanyHandle": "state-university-of-campinas" }, { "dateTo": "2010-07", "domain": "unicamp.br", "dateFrom": "2007-03", "jobTitle": "Scholarship Student", "description": "- Research on nonlinear systems modelling.<br>- Development of a forum and a blog for a linguistic students group.<br>- Design and programming of online graduate application system.", "linkedinCompanyHandle": "state-university-of-campinas" } ], "linkedinUrl": "https://www.linkedin.com/in/conradomiranda", "seniorities": [ "Director" ] }
I'd like the 3rd option as the others are too manual
Nope it doesn't work
What I'd like is Top-level fields
ID → id
First Name → firstName
Last Name → lastName
Name (Full) → name
Location → location
State → state
Country (2-letter) → country
Domain → domain
Job Title (Current) → jobTitle
Summary → summary
LinkedIn URL → linkedinUrl
Parent.*
Parent Org → parent.Org
Parent Title → parent.Title
Parent Country → parent.Country
Parent Headline → parent.Headline
Parent LinkedIn URL → parent["LinkedIn URL"] ← (use bracket notation for keys with spaces)
Parent Company (Experience) → parent["Company - Experience"]
Parent Job Title (People) → parent["Jobtitle - People"]
Parent Seniority (People) → parent["Seniorities - People"]
Parent Company Size → parent["Companysize - Company - People"]
Company.*
Company Name → company.name
Company Size → company.companySize
Company Logo URL → company.logo
Arrays (choose how you want to display)
Seniorities (list) → seniorities
If you want just the first value: seniorities[0]
Departments (list) → departments
First value: departments[0]
Skills (list) → skills
First value: skills[0]
Experiences (first/current role or a few key fields)
Exp 0 Title → experiences[0].jobTitle
Exp 0 Company Domain → experiences[0].domain
Exp 0 From → experiences[0].dateFrom
Exp 0 To → experiences[0].dateTo
Exp 0 LinkedIn Handle → experiences[0].linkedinCompanyHandle
Hi Ruben, thanks for reaching out! Have you considered using Sculptor (our chat with table) feature for this? I did this in my test account, where I prompted it and asked it to create separate columns from my Claygent results. This worked well for 3 of the columns ("Is London Based", "Clinic Website URL", and Has Counsellor Opening") but it forgot to add more columns for the other results - so it would probably take a few iterations of prompting to get all of the columns.
To access Sculptor, just click "Chat with Sculptor"! This will automate the creation of the columns, but don't think it will save you too much time when compared to manually clicking "Add as column". Let me know if you have further questions or need additional help! 😊
Natalie H. I don’t find a way to make this work Btw check the video from Yosh: https://www.clay.com/university/lesson/enrich-with-perplexity You’ll see he’s gotten a way to turn the answer into 3 columns but nor the video nor the text below explains how he’s managed this and I’ve tried and I can’t make it work for now
Hey Ruben! Thanks for sending the video over! So from Yash's video, he ran the first row --> and then pulled out the columns like "Key Point 0", "Key Point 1", and "Key Point 2". Then, he ran the rest of the rows - which auto-populated the corresponding values in the Key Points columns. This is great feedback about the turning JSON into columns feature, thank you! And let me know if you have further questions or need additional help! 😊
Natalie H. that’s exactly what I don’t get - HOW did he pull out the column like Key Point 0???
Thanks Natalie H. This I got But how did he manage to separate the Key Point 1 from the JSON as it’s pure text in his case no?
Look in the column Perplexity AI https://app.clay.com/workspaces/784835/workbooks/wb_0t318j2ztPqvMTdhAbw/tables/t_0t35k1jsxUsEKGrVnhH/views/gv_0t35k1jhKXFifs44rvz It’s a single response how can I turn each company under this as a column?
Unfortunately with the Perplexity AI enrichment we can't define outputs, like we can do while using "Use AI". If you were to use a different enrichment like "Use AI", you would be able to define the fields -- this is where we would put "Key Point 1" in the response and define the type as "Text", for example.
Ah I see
OK this makes sense
Basically the more I dig into Clay the more I realise it’s not able to do what I’d like it to do and would feel quite “logical”
100%
.png){kind=logo}
