Posted by Context graphs & AI Reasoning have caught recent imagination. But rarely people have talked about real world use cases – Applied AI Reasoning & what it takes to build commercial use cases – via Pvt AI Reasoning Lab. Here’s the reality execution path.
AI Reasoning – Category forming
AI reasoning startups have raised upto $300m early stage funding, so are world models based startups. But I cover Applied AI reasoning – which needs much frugal budget.
Some frontier model based startups – unicorns without product
But we cover capital efficient downstream Small reasoning models embedded in enterprise workflows
I cover much downsteam effects of building Small Reasoning Models, Data, Infra & application layer than the foundational model. This entire makes Decision Intelligence layer embedded in enterprise workflows.
Low-mid capex Applied AI reasoning & orchestration
Foundational models like Thinking Machine Labs or General Intuituion work on frontier AI. However Adept, Manus even to some extent Deepseek shows that it’s possible to work on Small Reasoning models, orchestration.
Small AI Reasoning model, orchestration – Vaayu’s work
We propose a pragmatic & well rounded path to exectution. This we have been slowly building. We are proposing an AI Pvt lab on Applied AI Reasoning. Here’s our journey from real execution view.
Our story building a ๐๐ซ๐ข๐ฏ๐๐ญ๐ ๐๐ ๐๐๐๐ฌ๐จ๐ง๐ข๐ง๐ ๐ฅ๐๐โand learning reality.
AI Reasoning embedded in enterprise workflow – example with Invoice Financing. 2-8% jump in profits.
AI reasoning, real use case MVP we built (2023):
AI reasoning, real use case (2023):
โIf profits jump 2โ8%, how will your stock react?โ – Achal Mehra asked in an automotive company’s boardroom. All via AI optimising enterprise workflows.
No new factory. No inventory. No sales cycle.
We were pitching dealer financing (where we started) with AI reasoning as the differentiator.
๐๐ฅ๐ฅ ๐ฌ๐๐๐ฆ๐๐ ๐ ๐จ๐ข๐ง๐ ๐จ๐ฎ๐ซ ๐ฐ๐๐ฒ.
Some well-wishers got us Fortune 200 intros.
Some leads came from 7-second cold pitches at Mumbai events.
Post quitting jobs, it felt big.
Angel round raised. Large customer demos. Many VC meetings.
๐๐ก๐๐ง ๐ซ๐๐๐ฅ๐ข๐ญ๐ฒ ๐ก๐ข๐ญ๐ฌ ๐ก๐๐ซ๐. They do for immediate outcomes.
Customers donโt pay for innovation. Investors donโt either.
So we said – Let’s slow AI, focu on distribution.
Plan: survive โ scale โ build deeper AI โ Something noone is building
๐๐๐๐จ๐ง๐ ๐ฉ๐ฎ๐ง๐๐ก ๐ข๐ง ๐๐๐๐:
Invoice financing collapsed post-ZIPR era.
I would wake up at 4am, code & survive on 6โ8 Punjabi-style large teas.
But we survived.
Brought AI back to the centre.
Shifted to Vaayu – AI for Finance & Sales. It was right move.
One hard lesson:
๐๐ง๐ง๐จ๐ฏ๐๐ญ๐ข๐จ๐ง ๐ฅ๐จ๐จ๐ค๐ฌ ๐ฌ๐ก๐ข๐ง๐ฒ ๐จ๐ง ๐ฌ๐จ๐๐ข๐๐ฅ ๐ฆ๐๐๐ข๐. ๐๐ฑ๐๐๐ฎ๐ญ๐ข๐จ๐ง ๐ง๐๐๐๐ฌ ๐ฆ๐๐ฌ๐ฌ๐ข๐ฏ๐ ๐๐ซ๐จ๐ง๐ญ-๐ฅ๐จ๐๐๐๐ ๐๐จ๐ง๐ฏ๐ข๐๐ญ๐ข๐จ๐ง.
Not because just cz tech is hardโbut because reference points are missing.
We been making constantly to build ๐๐ซ๐ข๐ฏ๐๐ญ๐ ๐ฅ๐๐ ๐จ๐ง ๐๐ ๐ซ๐๐๐ฌ๐จ๐ง๐ข๐ง๐ .
What we have tried & built, we will cover in next post.
Recent posts by Jaya Gupta & Ashu Garg on Context Graphs & Outcome-as-a-Service pioneered by Vaibhav Domkundwar, Gokul Rajaram’s view on vertical intelligence hint many more categories will emerge on the path to AI reasoning.
๐๐จ, I ๐ญ๐ก๐จ๐ฎ๐ ๐ก๐ญ ๐จ๐ ๐ฌ๐ก๐๐ซ๐ข๐ง๐ ๐จ๐ฎ๐ซ ๐๐๐ฆ๐จ ๐๐๐๐ค ๐๐ซ๐จ๐ฆ 2023.
(Now that we have built more stuff & use cases too)
But – I can tell for sure – there will be cycles of enthusaisim & great inflexion points. But execution path on AI reasoning, selling to customers, getting capital is equally hard.
๐๐๐ฏ๐ข๐ง๐ ๐ ๐๐๐ฌ๐ ๐ข๐ง ๐๐ & ๐ ๐ญ๐๐๐ฆ ๐ข๐ง ๐๐ง๐๐ข๐ ๐ข๐ฌ ๐ฉ๐ซ๐๐๐ญ๐ข๐๐๐ฅย for both of above to increase chances.
Then goal is – build realistic – keep looking for home runs.
If building in this space – happy to have chat.
Tagging people who mentored us OR have interest in area & innovation.
In next post – Building Applied AI Reasoning lab in capital efficient manner
If you prefer more techncial style & depth – pls request here.
If building in this space – happy to have chat.
Tagging people who mentored us OR have interest in area & innovation.
Read more – Beyond Context Graphs – AI Reasoning & World models