Ritvik Mandyam (@RitvikMandyam) header
Ritvik Mandyam (@RitvikMandyam) profile picture

Founder, Builder, High-Quality Dataset Enjoyer. Currently working on a startup that builds data analysis tools for humans.

Bangalore, KA, India
Joined August 2017
Photo by Ritvik Mandyam (@RitvikMandyam)Photo by Ritvik Mandyam (@RitvikMandyam)
Your org chart is missing a role: whoever gets the question after the dashboard runs out of answers. We call them the reluctant analyst. Our new post explores how this job forms, why more dashboards rarely fix it, and where self-service should stop. Full post in first reply.
1
1
1
20
The real test for an AI analyst isn't "can it answer the question?" It is "how quickly can a human challenge the answer and get to something defensible?"
9
We all talk about using semantic layers as a "single source of truth," but the real goal is to have something that makes data conversations boring. You never have to wonder whose definition of MRR got used for a calculation, because your semantic layer figured it out.
1
2
20
An analytics agent saying "I don't have enough information to answer that cleanly" isn't a failure. It's one of the best signs that the system might be safe to use when the answers need to be right.
1
9
The real test for an AI data answer isn't whether it sounds right in chat. It's what happens to the answer when someone asks, "Wait, whose definition of revenue is this?" If the system can't answer that, it's not governed - just articulate. Beauty's only truth to poets.
1
16
Text-to-SQL isn't the product. The product is knowing what the question means in your org, which source gets the final say, who is allowed to ask, what query ran, and who can inspect the answer later. SQL is the middle bit. The trust layer is the thing people buy.
5
Data people: where do you draw the line between: 1. this deserves a #dashboard forever 2. this deserves one verified answer now 3. please stop asking this question until we define the metric Feels like #AI #analytics only works if teams can tell those apart. #DataAnalytics
1
1
271
Your data is in 6 places. Your answers shouldn't be. Factal is the AI data analyst that works across your CRM, warehouse, product DB, and marketing tools - and actually respects your data governance. Ask. Get answers. Ship faster.
1
1
11
It's official - Factal's #GoogleAds integration is both smarter and more reliable than Google Ads Advisor. Plus Factal can do more than just run GAQL 😜
1
11
After two days of blood, sweat and cache misses, Factal's #GoogleAds integration is FINALLY smart enough to work with Perfomance Max campaigns. SO glad we can use our own tool on our own data again.
1
9
Just used the #GoogleAds integration in Factal.io to figure out which of our assets were being shown to people most often. I don't even know which PAGE that table is on in the Ads UI - Factal told me in under 30s.
11
#ChatGPT is probably the most advanced #AI humans have created, and it's made up of fewer neurons than an earthworm's brain. Talking to ChatGPT is like asking an earthworm its opinions on life.
51
Don't you love that moment when you find five Stack overflow questions with your exact bug, and none of them have any answers? Yeah, me neither.
I asked the IT department at my university to help me figure out the IP address of my Raspberry Pi. Nothing suspicious about a student trying to find out the IP address of a computer that he doesn't have access to, nothing at all.
1
1
#DeepMind just used #MachineLearning to improve matrix multiplication algorithms, which they will now use to improve #MachineLearning algorithms.
Apple's M series of chips: their latest attempt to make sure that if you bought Apple, you stay Apple.