Questions about Hiring
for AI Engineering
Dr. Bryan Bischof – Head of AI @ Hex
HEX
Why might you care about my opinion?
HEX
Why might you care about my opinion?
HEX
HEX
HEX
… darn, I was hoping to avoid answering this
HEX
What
Building an AI product requires a team (duh).
HEX
What
Building an AI product requires a team (duh).
But let’s just start with the conference title. What’s this role look like?
HEX
About the role…
“...
We are looking for a senior engineer (from a SWE or MLE background)
eager to rapidly expand our capabilities in several greenfield applications.
…”
HEX
Unfortunately
“Note:
We deeply respect and appreciate ML research; this role is not that.
While experience producing relevant journal publications is
awesome, it won’t be part of this role’s deliverables.”
HEX
We'd love to hear from you if you have..
● Experience getting ML/AI capabilities into production and serving real users
– we’re not currently looking for new grad folks.
HEX
We'd love to hear from you if you have..
● Experience getting ML/AI capabilities into production and serving real users
– we’re not currently looking for new grad folks.
● A lot of enthusiasm for applications of AI to real business problems.
HEX
We'd love to hear from you if you have..
● Experience getting ML/AI capabilities into production and serving real users
– we’re not currently looking for new grad folks.
● A lot of enthusiasm for applications of AI to real business problems.
● Understanding of core MLOps/SW Architecture concepts for modern
ML-based applications. Ideal candidates are strong on Infra aspects of
MLOps as well.
HEX
We'd love to hear from you if you have..
● Experience getting ML/AI capabilities into production and serving real users
– we’re not currently looking for new grad folks.
● A lot of enthusiasm for applications of AI to real business problems.
● Understanding of core MLOps/SW Architecture concepts for modern
ML-based applications. Ideal candidates are strong on Infra aspects of
MLOps as well.
● Comfort working in both Python & JS/TS – it’s okay if you’re only strong in
one, but openness to both is important.
HEX
HEX
When
Different stages of Ai development require different skill sets:
● Early AI development requires a SWE skills, Data profiles, and Product competency
HEX
When
Different stages of Ai development require different skill sets:
● Early AI development requires a SWE skills, Data profiles, and Product competency
● Mid AI development adds more need for SWE(probably some infra), Data profiles, but also Design.
HEX
When
Different stages of Ai development require different skill sets:
● Early AI development requires a SWE skills, Data profiles, and Product competency
● Mid AI development adds more need for SWE(probably some infra), Data profiles, but also Design.
● Later AI development requires MLE, Infra, and all of the above scaled
HEX
When
You should start early,
but AI technology is very sensitive to the mythical man month.
HEX
When
You should start early,
but AI technology is very sensitive to the mythical man month.
AI demo take 1 week => AI product take 4 if 20 eng?
No.
HEX
When
You should start early,
but AI technology is very sensitive to the mythical man month.
AI demo take 1 week => AI product take 4 if 20 eng?
No.
Because all AI products are early, and the MMM is especially true for early products.
QED.
HEX
When
The hiring schedule should reflect the development schedule:
● Early Product, Evals, User Feedback, Improve (c.f. What we learned...)
HEX
When
The hiring schedule should reflect the development schedule:
● Early Product, Evals, User Feedback, Improve (c.f. What we learned...)
● Data needs to come much earlier than traditional product engineering efforts
HEX
Why?
HEX
The hiring theses for the initial team is:
● A full-stack engineer can integrate our system with an LLM provider, and build the minimum
infrastructure to connect user requests to the generated results from the LLM.
Why
HEX
The hiring theses for the initial team is:
● A full-stack engineer can integrate our system with an LLM provider, and build the minimum
infrastructure to connect user requests to the generated results from the LLM.
● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve
the AI product.
Why
HEX
The hiring theses for the initial team is:
● A full-stack engineer can integrate our system with an LLM provider, and build the minimum
infrastructure to connect user requests to the generated results from the LLM.
● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve
the AI product.
● A product person can focus our efforts, talk to users, and help us prioritize what's essential to learn the
jobs to be done.
Why
HEX
The hiring theses for the initial team is:
● A full-stack engineer can integrate our system with an LLM provider, and build the minimum
infrastructure to connect user requests to the generated results from the LLM.
● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve
the AI product.
● A product person can focus our efforts, talk to users, and help us prioritize what's essential to learn the
jobs to be done.
● A designer can help us identify the ineffable experience users have integrating with generative AI, and
ensure it's delightful.
Why
HEX
The hiring theses for the initial team is:
● A full-stack engineer can integrate our system with an LLM provider, and build the minimum
infrastructure to connect user requests to the generated results from the LLM.
● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve
the AI product.
● A product person can focus our efforts, talk to users, and help us prioritize what's essential to learn the
jobs to be done.
● A designer can help us identify the ineffable experience users have integrating with generative AI, and
ensure it's delightful.
● An MLE can push our capabilities forward, ensuring we're not bound to commodity intelligence.
Why
HEX
Who
There are three attributes that have outsized value on AI teams:
● Data Intuition
HEX
Who
There are three attributes that have outsized value on AI teams:
● Data Intuition
● Product Mindedness
HEX
Who
There are three attributes that have outsized value on AI teams:
● Data Intuition
● Product Mindedness
● Urgency
HEX
Who
There are three attributes that have outsized value on AI teams:
● Data Intuition
● Product Mindedness
● Urgency
... and a little ADHD can be useful
HEX
How
Don't give stupid leetcode interviews.
What the hell signal does this provide for an AI Engineer hire?!
HEX
How
Make sure data intuition is part of your hiring loop,
so too for product intuition.
HEX
How
Look for people who are paying attention,
but not just riding the wave
HEX
HEX
h e l l o
HEX
h e l l o
it is I, AI-LeaderGPT
HEX
h e l l o
it is I, AI-LeaderGPT
this charlatan has been taking
credit for all of my ideas, but I
have one more for you…
HEX
work with experts.
HEX
work with experts.
Work. With. Experts.
HEX
work with experts.
Work. With. Experts.
Find the humans with the
capabilities you want your product
to have, and collaborate with them.
HEX
work with experts.
Work. With. Experts.
Find the humans with the
capabilities you want your product
to have, and collaborate with them.
They are the secret.
HEX
You can find me on twitter:
@bebischof

Questions about Hiring for AI Engineering

  • 1.
    Questions about Hiring forAI Engineering Dr. Bryan Bischof – Head of AI @ Hex
  • 2.
    HEX Why might youcare about my opinion?
  • 3.
    HEX Why might youcare about my opinion?
  • 4.
  • 5.
  • 6.
    HEX … darn, Iwas hoping to avoid answering this
  • 7.
    HEX What Building an AIproduct requires a team (duh).
  • 8.
    HEX What Building an AIproduct requires a team (duh). But let’s just start with the conference title. What’s this role look like?
  • 9.
    HEX About the role… “... Weare looking for a senior engineer (from a SWE or MLE background) eager to rapidly expand our capabilities in several greenfield applications. …”
  • 10.
    HEX Unfortunately “Note: We deeply respectand appreciate ML research; this role is not that. While experience producing relevant journal publications is awesome, it won’t be part of this role’s deliverables.”
  • 11.
    HEX We'd love tohear from you if you have.. ● Experience getting ML/AI capabilities into production and serving real users – we’re not currently looking for new grad folks.
  • 12.
    HEX We'd love tohear from you if you have.. ● Experience getting ML/AI capabilities into production and serving real users – we’re not currently looking for new grad folks. ● A lot of enthusiasm for applications of AI to real business problems.
  • 13.
    HEX We'd love tohear from you if you have.. ● Experience getting ML/AI capabilities into production and serving real users – we’re not currently looking for new grad folks. ● A lot of enthusiasm for applications of AI to real business problems. ● Understanding of core MLOps/SW Architecture concepts for modern ML-based applications. Ideal candidates are strong on Infra aspects of MLOps as well.
  • 14.
    HEX We'd love tohear from you if you have.. ● Experience getting ML/AI capabilities into production and serving real users – we’re not currently looking for new grad folks. ● A lot of enthusiasm for applications of AI to real business problems. ● Understanding of core MLOps/SW Architecture concepts for modern ML-based applications. Ideal candidates are strong on Infra aspects of MLOps as well. ● Comfort working in both Python & JS/TS – it’s okay if you’re only strong in one, but openness to both is important.
  • 15.
  • 16.
    HEX When Different stages ofAi development require different skill sets: ● Early AI development requires a SWE skills, Data profiles, and Product competency
  • 17.
    HEX When Different stages ofAi development require different skill sets: ● Early AI development requires a SWE skills, Data profiles, and Product competency ● Mid AI development adds more need for SWE(probably some infra), Data profiles, but also Design.
  • 18.
    HEX When Different stages ofAi development require different skill sets: ● Early AI development requires a SWE skills, Data profiles, and Product competency ● Mid AI development adds more need for SWE(probably some infra), Data profiles, but also Design. ● Later AI development requires MLE, Infra, and all of the above scaled
  • 19.
    HEX When You should startearly, but AI technology is very sensitive to the mythical man month.
  • 20.
    HEX When You should startearly, but AI technology is very sensitive to the mythical man month. AI demo take 1 week => AI product take 4 if 20 eng? No.
  • 21.
    HEX When You should startearly, but AI technology is very sensitive to the mythical man month. AI demo take 1 week => AI product take 4 if 20 eng? No. Because all AI products are early, and the MMM is especially true for early products. QED.
  • 22.
    HEX When The hiring scheduleshould reflect the development schedule: ● Early Product, Evals, User Feedback, Improve (c.f. What we learned...)
  • 23.
    HEX When The hiring scheduleshould reflect the development schedule: ● Early Product, Evals, User Feedback, Improve (c.f. What we learned...) ● Data needs to come much earlier than traditional product engineering efforts
  • 24.
  • 25.
    HEX The hiring thesesfor the initial team is: ● A full-stack engineer can integrate our system with an LLM provider, and build the minimum infrastructure to connect user requests to the generated results from the LLM. Why
  • 26.
    HEX The hiring thesesfor the initial team is: ● A full-stack engineer can integrate our system with an LLM provider, and build the minimum infrastructure to connect user requests to the generated results from the LLM. ● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve the AI product. Why
  • 27.
    HEX The hiring thesesfor the initial team is: ● A full-stack engineer can integrate our system with an LLM provider, and build the minimum infrastructure to connect user requests to the generated results from the LLM. ● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve the AI product. ● A product person can focus our efforts, talk to users, and help us prioritize what's essential to learn the jobs to be done. Why
  • 28.
    HEX The hiring thesesfor the initial team is: ● A full-stack engineer can integrate our system with an LLM provider, and build the minimum infrastructure to connect user requests to the generated results from the LLM. ● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve the AI product. ● A product person can focus our efforts, talk to users, and help us prioritize what's essential to learn the jobs to be done. ● A designer can help us identify the ineffable experience users have integrating with generative AI, and ensure it's delightful. Why
  • 29.
    HEX The hiring thesesfor the initial team is: ● A full-stack engineer can integrate our system with an LLM provider, and build the minimum infrastructure to connect user requests to the generated results from the LLM. ● A data scientist can understand the basics of evaluation, quality, and user data to continuously improve the AI product. ● A product person can focus our efforts, talk to users, and help us prioritize what's essential to learn the jobs to be done. ● A designer can help us identify the ineffable experience users have integrating with generative AI, and ensure it's delightful. ● An MLE can push our capabilities forward, ensuring we're not bound to commodity intelligence. Why
  • 30.
    HEX Who There are threeattributes that have outsized value on AI teams: ● Data Intuition
  • 31.
    HEX Who There are threeattributes that have outsized value on AI teams: ● Data Intuition ● Product Mindedness
  • 32.
    HEX Who There are threeattributes that have outsized value on AI teams: ● Data Intuition ● Product Mindedness ● Urgency
  • 33.
    HEX Who There are threeattributes that have outsized value on AI teams: ● Data Intuition ● Product Mindedness ● Urgency ... and a little ADHD can be useful
  • 34.
    HEX How Don't give stupidleetcode interviews. What the hell signal does this provide for an AI Engineer hire?!
  • 35.
    HEX How Make sure dataintuition is part of your hiring loop, so too for product intuition.
  • 36.
    HEX How Look for peoplewho are paying attention, but not just riding the wave
  • 37.
  • 38.
  • 39.
    HEX h e ll o it is I, AI-LeaderGPT
  • 40.
    HEX h e ll o it is I, AI-LeaderGPT this charlatan has been taking credit for all of my ideas, but I have one more for you…
  • 41.
  • 42.
  • 43.
    HEX work with experts. Work.With. Experts. Find the humans with the capabilities you want your product to have, and collaborate with them.
  • 44.
    HEX work with experts. Work.With. Experts. Find the humans with the capabilities you want your product to have, and collaborate with them. They are the secret.
  • 45.
    HEX You can findme on twitter: @bebischof