interviewed a Lead Data Scientist from a consulting firm
> no clue abt Coding Agents
> hasn't built anything in 2 yrs
> doesn't know how his own solutions were deployed
bro, AI is coming after you ! 💀
Senior folks should be building hands on at least 50 % time.
Hiring Full Stack AI Engineer !
> job is to build with AI
> talk to business, find a problem, build it yourself ( frontend, backend etc)
> Own everything
Requirements
> 1-2 yrs of AI Native experience
> Hunger to experiment with AI
DM your portfolio !
I literally keep saying this to everyone.
> as folks in tech ,we get too attached to the solution, how elegant /complex it is
> businesses cares about Outcome only !
what local businesses don't want:
- ai agents to chat with (creates work for them)
- skills for claude/gpt, etc.
- a knowledge graph / brain
- learning how to code & build internal tools
- automation with unclear business outcomes
what they do want:
- more calls, more customers, less headaches
keep the sauce as part of your internal process, sell outcomes and clarity. the fact that you use AI is a given
that is translated to the customer in terms of the value you create, the time you save them, and the price at which you deliver the outcomes
was getting into a comfort zone.
with AI moving so fast, this is the worst thing I can do to myself !
took some bold decisions. taking on some new challenges.
backing myself even more.
POV: you finish an AI Engineer interview and the candidate says :
“Best AI interview experience I’ve had till date ”
🫡
Candidate experience >>>
@hijunedkhatri got this shock when hiring in my team.
pure demand supply play + a completely opaque product.
most enterprises have complete dependency on Salesforce CRM and then feature upsell & maintenance cycle continues.
@kmeanskaran I follow a similar approach when I interview candidates.
understanding of what, why & how with ability to communicate matters.
you get to know about the buzzword candidates within the first 10 mins.
this is spot on !
the model capabilities and the harness are evolving fast.
but the basics remain same
> data
> documentation
because context is the most imp piece of the puzzle
interviewed few folks for AI Engineer Role today !
one thing stood out clearly -
> people who know 'WHY' differentiate themselves
> those who know the 'WHY' behind their design choices' and can explain it
#2 - AI Conversation Analysis
> Exploring @ElevenLabs STT
- Started with multi channel WAV
- Split into per-speaker mono files
Thought it'd save cost 🤦🏻♂️ Nope !
- Each file = separate API call → ~2× cost
Next challenge :
Stitching transcripts into turn-by-turn dialogue
Interviewed an AI Engineer
> 4+ yr of exp, impressive resume
> Projects across multimodal LLMs and Voice AI
> but struggled in explaining the problem statement and the approach
in interviews,
ability to communicate what problem you solved and how you solved it >>>> Tech
#1 - AI Conversation Analysis
> evaluating the right STT model
priority
- transcription accuracy
- Indic support (hindi, marathi)
- code switching (hinglish)
- speaker diarisation
latency not a concern
benchmarking @SarvamAI and @ElevenLabs for now !