Head of Ops & Senior AI PM @ Sarana AI

Building products with purpose,guided by data, grounded in empathy.

Hybrid Product Manager and UX designer shipping AI products that people actually use. I lead with curiosity, ship with rigor, and measure what matters.

See my workLet's chat
Scroll to explore
01 — About

I started in UX, fell in love with the why behind the pixels, and moved into product to ship things that actually matter.

Today I lead operations and AI product at Sarana AI — owning the messy intersection of strategy, research, and delivery. My approach: listen hard, validate with data, ship the smallest thing that learns, and iterate without ego.

  • 2+
    Years in Product
  • 1
    Year in UX Design
  • AI
    Product Focus
  • Ops
    Leadership
02 — Selected work

Four bets, four lessons, measurable outcomes.

Cutting AI agent hallucinations 62% with a human-in-the-loop review
01Case study

Cutting AI agent hallucinations 62% with a human-in-the-loop review

The problem

Customer-facing AI agent was confidently wrong on 1 in 4 replies, eroding trust in the first product week.

Role

Senior AI PM · led 4 engineers, 1 designer, 2 annotators

Approach

  • Ran 18 user interviews to map where trust broke down
  • Shipped a lightweight review queue before scaling the model
  • Instrumented agent confidence + reviewer overrides as first-class metrics
0%
Hallucination ↓
0/5
Trust score
0×
Activation lift

What I learned — Trust compounds slower than features. Slow the launch, ship the safety net first.

Rewriting onboarding to cut time-to-value from 12 min → 90 sec
02Case study

Rewriting onboarding to cut time-to-value from 12 min → 90 sec

The problem

70% of new users churned before activation. The product was strong; the first run wasn't.

Role

Product Manager · 2 engineers, 1 designer

Approach

  • Mapped the full activation funnel with session replays
  • Cut 11 onboarding screens down to 3 progressive disclosures
  • A/B tested two empty-state strategies before committing
0%
Activation rate
0%
Time saved
0%
D7 retention ↑

What I learned — Onboarding isn't a screen — it's the moment the user feels capable.

Turning a noisy analytics dashboard into a decision tool
03Case study

Turning a noisy analytics dashboard into a decision tool

The problem

Internal ops team had 40+ metrics and zero clarity on what to act on each Monday.

Role

Head of Ops · cross-functional with data, RevOps, CX

Approach

  • Interviewed 9 internal stakeholders about their actual Monday rituals
  • Defined a 3-tier metric hierarchy: pulse, signal, diagnostic
  • Replaced the dashboard with a weekly narrative + 4 north-star tiles
0%
Faster reviews
0/9
Stakeholders aligned
0×
Decisions/week

What I learned — If a metric doesn't change someone's behavior, it shouldn't be on the dashboard.

Designing a collaboration ritual that scaled with a 3× headcount
04Case study

Designing a collaboration ritual that scaled with a 3× headcount

The problem

Eng, design, and ops were drifting as Sarana AI tripled in size. Roadmaps existed; alignment didn't.

Role

Ops & PM lead · org-wide rollout

Approach

  • Audited every recurring meeting; killed 40% of them
  • Introduced async product briefs + a weekly demo cadence
  • Made roadmap context one click away in every Slack channel
0%
Fewer meetings
0%
Team NPS
0 wk
To full alignment

What I learned — Process is just empathy with a calendar invite. Design it, don't inherit it.

03 — Toolbox

The craft, the tools, the muscles.

Strategy & Research

  • Product Strategy
  • User Research
  • Jobs-to-be-Done
  • Roadmapping

Design & Prototyping

  • Figma
  • Wireframing
  • Design Systems
  • UX Writing

Data & Experimentation

  • A/B Testing
  • Mixpanel
  • Amplitude
  • SQL

AI & Delivery

  • LLM Product Design
  • Eval Frameworks
  • Agile
  • Linear
04 — How I work

A five-step loop I've refined on every product I've shipped.

  1. I run lean discovery sprints: 8–12 interviews, jobs-to-be-done mapping, and a hypothesis board. Every project starts with a clear problem statement someone outside product could repeat back.

  2. Tight one-page PRDs, a single success metric, and explicit non-goals. If the team can't explain the bet in one sentence, we keep cutting.

  3. Bias toward thin slices and dogfooding. I work shoulder-to-shoulder with eng + design, unblock daily, and protect the team from scope creep that doesn't change the metric.

  4. Every release ships with the analytics and eval queries that will tell us if it worked. Weekly metric reviews are non-negotiable; surprises are the enemy.

  5. Retros that name what we got wrong, written post-mortems for shipped bets, and a public changelog of what we'd do differently. Compounding wins come from compounding lessons.

05 — Kind words
“
Indira is the rare PM who can hold strategy and craft in the same hand. She made our roadmap legible and our team braver.
Engineering Lead · Sarana AI
“
She turned a tangled AI workflow into something users actually trust. Empathy in, evidence out — that's how she ships.
Design Partner · Former collaborator
06 — Let's talk

Let's build something
that matters.

Drop a note — I reply within 24 hours. Coffee, collaboration, candid product chats, all welcome.

indirasetia28@gmail.comLinkedIn