Sign In

Dhruv Guliani

Building next-gen customer simulations @ Keplar ⚡We’re Hiring!

Dhruv Guliani is a Mechatronics Engineer with a background in software and embedded product development, currently specializing in Machine Learning and Artificial Intelligence. His technical expertise includes Python (TensorFlow, Keras, Scikit-learn), C++/C, and JavaScript (Node.js). On the non-technical side, he excels in leading small, cross-functional teams, strong written and verbal communication, and navigating through ambiguity. Dhruv enjoys connecting with ambitious individuals tackling challenging and meaningful problems and welcomes conversations on these topics.

Dhruv attended the University of Waterloo, where he pursued a Bachelor of Applied Science (B.A.Sc.) degree in Mechatronics, Robotics, and Automation Engineering. He also completed his International Baccalaureate Diploma at Turner Fenton Secondary School.

His professional journey includes roles at reputable organizations such as Keplar.io, South Park Commons, Google, Connected, Kindred.ai, Apple, Connected Lab, Tesla Motors, BlackBerry, and Future Shop. His experiences range from co-founder and founder fellow to software engineer, artificial intelligence engineer, firmware engineer, and various software development positions. Dhruv's diverse background demonstrates his adaptability and proficiency in multiple areas within the tech industry.

Highlights

May 3 · linkedin

💡Builder insight #3: Is language a substrate for intelligence?

📖 In the 60’s, Noam Chomsky came up with the theory of Universal Grammar, the idea that the human brain has an innate structural ability to learn language.

🧠 In other words, the structure of language implicitly encodes how our brains’ perceive, think, reason, and decide. This would mean that modeling language is the first step to modeling the human mind in a practical way.

While we don’t know if the theory of Universal Grammar is right, we do see evidence to believe that LLMs create a substrate, a sort of medium, for higher order intelligence, reasoning, and learning to form. This is because language lets us encode a flexible range of higher order data and tasks concisely.

It moves us beyond binary.

🧐 However, using language is not so straight-forward and comes with its own challenges.  It’s lossy, imprecise, and context-dependent. Language has many ways to express similar things, and is wildly open to interpretation. This sort of imprecision and “looseness” can create a lot of challenges for builders, especially when trying to scale solutions (think reliability and latency, and the meaning of quality).

-> TLDR; prompting is harder than it looks, quality evaluation feels fuzzy

But, it also means that there is a huge opportunity to marry the flexible encoding ability of language with structure (through structured inputs/outputs, carefully curated calibration datasets, structured evals, etc.) to solve complex problems within the language domain.

For us, these complex and newly solvable problems include predicting how people will perceive stimuli, how people will react and respond to creative assets, and how economic decisions are made, etc. It’s a wild time to be alive! ☄️

Aug 31 · linkedin

Today we move the 🧀.

We're thrilled to announce keplar.io out of stealth! 🎉

Keplar is the first product to offer data-driven, hyper-realistic customer simulations that enable consumer companies to develop powerful test & learn flywheels. With Keplar, brands can minimize assumptions, deploy creatives faster than ever, and move at enterprise scale with startup speed.

Through countless conversations with brand and marketing managers, we learned that reach can be bought, but a killer creative or product strategy is impossible to brute force. It takes painstakingly slow test & learn cycles, either using traditional market research tools, agencies, single-use-case ML models or A/B testing.

💡So William and I asked ourselves whether it was possible for companies to combine modern ML advancements with rich public and first-party data troves to build hyper-realistic, evolving models of their customers.

What if these models were deeply rooted in psychology and behavioral science?  How might these “synthetic customers” augment existing R&D methods to create a truly iterative creative process?  How might they speed up the development of marketing assets, product concepts, and brand strategy?

⚡These questions got us excited and led us to Keplar, the world’s first customer simulation platform (CSP)!

If you’re building a brand, running a complex creative process for marketing, expanding your influencer program, or running surveys and focus groups, you won’t want to 💤 on this one!

Join us in building an ambitious and customer-led future.

Get early access: https://lnkd.in/g9JzMSE2 :)

#AI #marketing #brandstrategy #insights #innovation #cpg #fashion #testandlearn #keplar

Dhruv Guliani
Add to my network

Location

San Francisco, California, United States