Malika is devoted to bringing great ideas to life. She is an Operations Partner at Comet Labs, a cross between a venture fund and experimental research lab that supports AI and robotics startups. She previously worked in investment banking, and oversaw the development and growth of software and hardware startups in the education, healthcare, and telecom fields, in Asia, Europe and North America. She graduated from the University of Cambridge and has an MBA from Tsinghua University and MIT Sloan.
Malika Cantor, Operations Partner, Comet Labs at The AI Conference 2017
A VC’s Perspective on AI
AI CONFERENCE, JUNE 2017
0.5min: A bit about Comet
0.5min: Reality vs. Hype
1min: Does traditional VC work for AI startups?
3min: What AI startups need to succeed
5min: Insights into our Due Diligence process
➔ 2 years old
➔ Focused exclusively on AI Enabling Tech & Applications of AI
➔ The most active early-Stage investor in B2B AI and robotics.
➔ Fund + Labs model
➔ China + US dual perspective
40 + portfolio companies, follow on investment from:
The AI Revolution is here, despite the Hype
The scale and scope of change is larger than ever
Previous waves of
disruption are the
foundation of the next
has reached an
inflection point enabling
a new generation of
100x more touch-points
with technology creates
opportunities for more
intuitive products and
new business models
Traditional investment methods don't
work for AI startups
Meaningful solutions require
deep involvement of the
customers early on, as well as an
understanding of potential
markets and pain points.
The access to industry for any single
startup is extremely limited and
inefficient. Entrepreneurs need to find
pilot projects, early customers, and
tools for scaling.
What do AI startups need?
DD: Data Moat
● How are you obtaining enough data initially to provide satisfactory solutions
to your early customers? (Cold Start problem)
● Are you generating or negotiating access to proprietary data sets?
● If you are generating proprietary data sets:
○ What makes your data valuable and hard to obtain?
○ Do you own the data that you are collecting? Do you own it exclusively?
● If you are negotiating access to proprietary data set:
○ How did you negotiate access?
○ Exclusive license? Nonexclusive license?
● Walk us through the product technology stack, from data ingestion to output
○ Which part of the stack is hardest to replicate, and why?
○ What is your technical secret sauce?
○ Which parts did you build, and which parts are off the shelf?
■ Which libraries did you use?
● How much of a “human in the loop” is required, and where?
● Customer references
● At least one technical founder
● Who’s your first hire after the fundraise?
● Why are you the best team to work on this problem?
● What do you understand about the industry that others don’t?
● If you could rethink the industry from first principles, how would you do it?
● What is the immediately addressable market?
● Why now? Risk of being too early and running out of cash before mass
● What competitor do you most often run into when closing sales, and how
are you differentiated?
● Sales cycle? Payback period? Churn?
● Monthly recurring revenue?
● Who are your biggest customers?
● How have you been acquiring customers so far?
Questions? Email firstname.lastname@example.org :)