Product Management · AI Features

Building products
people feel.

I'm a PM focused on AI-powered experiences — turning complex technology into features that are intuitive, delightful, and measurably better for users.

AI / ML Products 0→1 Feature Work Growth & Retention B2C · Consumer
Spotify AI Feature Proposal

Mood Journey

An adaptive listening engine that evolves music in real time — matching energy, emotion, and context as your day unfolds.

–22%
Skip rate target
+18 min
Daily listening uplift
+11%
Artist discovery rate
D30 +8%
Retention uplift
The Problem
🎵
Playlists are frozen in time
Your "Chill Morning" mix doesn't know you're now three coffees deep and sprinting to a deadline.
😶
Skip fatigue is real
Users skip 30% of songs — not because they dislike the track, but because it doesn't fit the moment.
🔄
Context shifts constantly
Commute → focus work → gym → wind-down. One playlist cannot serve all four moods.
📉
Discovery drops off
Users stick to 10–15 familiar songs on repeat, limiting exposure to Spotify's 100M+ track catalog.
User Flow
1
Mood check-in (optional)
User opens Spotify. A subtle card asks "How are you feeling?" — or the system infers from time, calendar, and listening history. No mandatory friction.
2
Mood Journey session starts
A dynamic queue is generated — a living arc that evolves across 30 / 60 / 90 min. Opening tracks match current mood.
3
Real-time signal collection
Skip behavior, listening duration, time of day, and BPM engagement feed a continuous feedback loop. Model recalibrates every ~5 tracks.
4
Mood arc transitions
Music shifts in tempo and valence — from focus-mode instrumentals → mid-energy pop → calming lo-fi as evening arrives.
5
Journey recap + discovery card
Session summary shows energy evolution and surfaces 3 new artists aligned to the user's emotional arc. CTA: Save as playlist.
Feature Spec
Core engine Multimodal ML model trained on audio features (tempo, valence, energy, danceability), listening sequences, skip patterns, and contextual signals (time, calendar, weather via opt-in).
Signals Skip rate Listen duration Time of day Calendar (opt-in) BPM trajectory Weather API
User control Mood arc slider (Energize → Maintain → Wind down). Manual override at any point. Per-session memory toggle for privacy.
Platforms iOS, Android (v1). Desktop + Car View (v2). Full parity in Premium. Lite version for Free tier.
Privacy Mood inference runs on-device where possible. No raw audio captured. Session data not retained unless user opts into Cross-Session Memory.
Edge cases Cold-start: defaults to genre selection. Multi-user households: per-profile isolation. Offline: fallback to downloaded tracks closest to last detected mood.
Success Metrics
–22%
Skip rate during sessions
+18 min
Avg daily listening (Premium)
+11%
Artist discovery rate / month
40%
Sessions → "Save playlist" CTA
D30 +8%
Retention vs control group
4.4★
Feature CSAT target
Guardrails
Skip rate increases vs baseline — model may be over-correcting mood arcs
Uptick in "Turn off mood tracking" — indicates trust or transparency concern
Feature opt-in targets 35% of MAU within 90 days of launch
Launch Roadmap
Q1
Internal alpha — signal pipeline + model v1
Q2
Closed beta — 5k Premium users, 3 markets
Q3
Global Premium rollout + calendar integration
Q4
Free tier lite + social sharing of Mood Journeys
Duolingo
AI Streak Coach
Adaptive motivation engine that personalizes nudges based on learning velocity and streak risk signals.
Coming soon
Amazon
Intent-Aware Search
Multimodal search that infers purchase intent from browsing context, past orders, and seasonal signals.
Coming soon
Contact

Let's connect.

Open to PM roles, collaborations, and conversations about AI products.

✉
Email
kdhanurkar@gmail.com
→
✆
Phone
425-445-3495
→
in
LinkedIn
linkedin.com/in/khushidhanurkar
→