Product Manager · Applied AI · Connected Systems

Hi, I'm Yamna.
AI Product Manager.

I turn complex AI and connected systems into useful customer experiences, from discovery and requirements through launch and adoption.

Shipped on Mobile apps · Desktop and web · Enterprise systems · Wearables and VR · In-store hardware

ExploreInteractive demos · AI builds · Wireframe

Yamna Irfan at her laptop with headphones on

I built this site with Claude. Poke around, there's a lot to explore. Hope you have fun exploring it just as much as I had fun building it.

Highlights
About

Product and people.

Full-lifecycle technical product manager.

I care about understanding what people need, why an experience feels difficult, and what would make it more useful.

I combine technical fluency with customer and commercial judgment, keeping the people using a product at the center of the decisions behind it.

I grew up in Dubai before moving to Chicago, which taught me early that the same product can land very differently depending on who's using it. Studying marketing and psychology shaped how I think about behavior, friction, and trust.

People first, always. That means empathy for customers, kindness with teammates, and some grace for myself too.

Outside of product, I write comedy skits and release original music.

Experience
4 years in PM / 9 years across product, marketing and customer experience
Education
B.A. Integrated Marketing Communications / Psychology minor / Roosevelt University
Study abroad: Anglo-American University, Prague
Location
Chicago / Open to relocation
Experience

Where I've worked.

The short version. Click any role to jump to the full story.

  1. Meta Reality LabsProgram Manager, Product, AI & VRConnected voice AI quality with customer discovery, engagement, experimentation, and global go-to-market across two devices.Read the story →
  2. Coates GroupDelivery Lead, ProductAdapted recognition software to a new enterprise client's requirements, led a phased rollout, and improved release reliability across the platform.Read the story →
  3. CURO FinancialProduct ManagerTranslated lending logic and system dependencies into customer journeys, clear requirements, and aligned delivery priorities.Read the story →
  4. Purple Peacock StudiosCo-Founder | Product, GrowthOwn customer discovery, positioning, pricing, acquisition, service design, and commercial decisions.Read the story →
Work · Case studies
MetaVoice AI

From voice AI capability to customer adoption.

A voice feature has to do more than recognize a command. People need to discover it, trust the response, and have a reason to use it again. At Meta Reality Labs, I worked on Voice Command and AI Assistant experiences for Quest and Ray-Ban Stories, connecting model performance and usage patterns with product decisions and go-to-market.

65%Contributed to a 65% increase in Voice Command adoption through team launches.
Role
Program Manager, Product, AI & VR
Product
Voice Command, hands-free voice control
Surfaces
Quest headset, Ray-Ban Stories glasses, companion app, notifications, email, in-app messaging
I owned

Requirements, user stories, acceptance criteria and backlog priorities; engagement features across the app, notifications, email and in-app messaging; launch readiness.

With the team

Success metrics, experiments and post-launch analysis; speech and intent evaluation with ML Engineering and Data Science; go-to-market across two devices. Engineering and ML partners built the features and models.

Key product decisions
  • Whole-task success. Judged quality from spoken request to intended action to user feedback, not on model accuracy alone.
  • Discovery vs. fallback. Used usage patterns to separate experiences that needed clearer discovery from those that needed more reliable recognition and fallback behavior.
  • Across two devices. Balanced capability, latency, and expectations on a headset and on glasses with no screen, and planned launch and adoption with the team.
Product discoveryRequirements & backlogExperimentation & KPIsGo-to-marketAI evaluationCross-functional influence
When the model isn't sureA simplified simulation I made, not real Meta data

This shows the decision I worked on: when the device should act, double-check, or help the person try again. It plays through each situation on the headset, then the glasses. Pause anytime to pick a situation or move the slider.

Meta Quest · 1 of 4
Try it on
1 Turn it on in the app
Companion app
DeviceMeta Quest
Wake word"Hey Meta"
2 Say a command
noise
“Take a photo.”
3 What happens
Act
Photo taken.
Shutter sound and the capture light turns on.
The person says

“Take a photo.”

Quiet room, common phrasing.

What the system hears
Speech recognition94%
Intent recognition92%
Overall confidence92%

That small line is the threshold. Above it, the device just acts.

What the product does
  1. ActConfident: just do it.
  2. ConfirmClose: check before acting.
  3. Fall back and teachUnsure: help them succeed next time.
Photo taken.
Risk of a wrong action
Friction for the person

What I learned: a smarter model is only half the answer. The bigger decision was where to put that line, and what the device says on each side of it. I came to see the "fall back and teach" moment as one of the best chances to show people what they can say, instead of treating it as an error.

Read the full storyHide the full story

01Discovery

Adoption was the real problem: people never learned which commands existed, or tried one, got a miss, and stopped. I used command-usage patterns, behavioral signals, and speech and intent-recognition evaluation to see where people got stuck, and ran dogfooding cycles alongside the data.

WithData Science, UX, Research

02Product & model tradeoffs

Every command passes through speech recognition, intent recognition, and a device or app action, and each step fails differently across voices, accents, rooms, and hardware. With ML Engineering and Data Science, I separated product problems from model problems, using the decisions summarized above.

WithML Engineering, Data Science, UX

03Requirements

I translated those decisions into requirements, 100+ user stories, acceptance criteria, and backlog and sprint priorities, including how the device responds when it isn't sure what it heard.

WithEngineering, ML Engineering

04Collaboration

I ran the PM team's Agile process for a 15+ person cross-functional team, from sprint planning through UAT, working day to day with Engineering, ML Engineering, UX, Research, and QA.

WithEngineering, ML Engineering, UX, Research, QA

05Global GTM

I helped plan go-to-market with the team for large-scale features across Quest and Ray-Ban Stories globally, and aligned partners around launch readiness. I led engagement features across the mobile app, in-device notifications, email, and in-app messaging to help 5,000+ users find commands worth repeating.

WithProduct Marketing, Legal, Privacy, QA, Engineering

06Adoption

Experimentation and post-launch analysis helped us assess discovery, engagement, repeat usage, and feature performance. Result: contributed to a 65% increase in Voice Command adoption through team launches.

Coates GroupEnterprise software

Adapting recognition software for a new enterprise client.

During COVID, long drive-thru lines put pressure on Dunkin', a Coates Group client, and on its store teams. Coates had license plate recognition software originally built for another client. My work was adapting it to Dunkin's operational, legal, and product requirements and connecting it to the systems already running in their locations.

50 locations live, with demand for broader expansion
Employer
Coates Group
Client
Dunkin'
Role
Delivery Lead, Product
Product
License plate recognition and digital menu software
I owned

Rollout coordination across development, design, QA, installation, integrations, and production support. Separately, I led the team's introduction of a new release process for the Switchboard CMS.

With the team

Adapting requirements for a new client and the phased integration strategy, with Engineering, Data Science, UX Research, and Dunkin' teams, who built and installed it.

Also measured

Release cycles moved from about monthly to about every two weeks after the CMS release change.

Key product decisions
  • Adapt, don't rebuild. Adapted an existing recognition capability to Dunkin's product, operational, and legal requirements.
  • Phased rollout. Coordinated a bounded first rollout across 50 locations before broader expansion.
  • Release reliability, separately. Led the team's introduction of a gated release process for the Switchboard CMS.
Customer requirementsProduct tradeoffsSystems integrationPhased rolloutRelease managementCross-functional delivery
Explore the drive-thru journeyA sample I made. Timings and values are illustrative, not real Coates or Dunkin' data

Follow a car through the drive-thru. See what the customer sees on the digital menu board and how recognition, ordering, and store systems work together behind the scenes.

08:42:10
A car drives through a drive-thru: past a license plate recognition camera, a digital menu board, an order speaker, and the pickup window. PICK UP Coates kiosk · connected to LPR technology Good morningSeasonal latte Welcome backYour usual? Order #1042Total $5.18 Thank you!See you tomorrow Plate KTM 3719 · 97% Guest matched Menu personalized Sent to POS Order handed off KTM 3719
Good morningLane 1
SeasonalMaple oat latteAsk for it at the speaker
Welcome backYour usual?Medium iced coffee, oat milkEverything bagel, toastedJust say "my usual"
Order #1042
Medium iced coffee, oat milk$3.19
Everything bagel, toasted$1.99
Total$5.18
Please pull forward
Thank you!See you tomorrow
Hot coffee$2.49
Iced coffee$3.19
Latte$3.89
Bagel$1.99
Breakfast sandwich$4.49
Donuts, 6 pack$7.99
What the customer sees · digital menu board
Store operationsSimulated
License plate
recognition
Order
history
Switchboard
CMS
Point of
sale
Event log0 events
    Plate--
    Confidence--
    Time in lane0:00
    Read the full storyHide the full story

    01Client needs

    There were two customers to keep in view: Dunkin' as the business client, and the people waiting in its drive-thru lines. I worked with the team to understand how Dunkin's customer workflows, operations, legal requirements, and systems differed from the original implementation.

    WithDunkin' client teams, UX Research

    02Adapting requirements

    The recognition capability was built for another client with different product and legal requirements. We adapted it instead of rebuilding it, and I translated the differences into delivery priorities, balancing customer value with limited resources, pandemic-era hardware constraints, and the reliability live locations need.

    WithEngineering, Data Science, UX Research

    03Integrations

    Recognition had to work with customer order history, the Switchboard CMS, POS systems, and drive-thru hardware, alongside crew workflows. I coordinated development, design, QA, installation, and integrations across those dependencies.

    WithEngineering, design, QA, installation

    04Phased rollout

    We started with 50 locations: a bounded launch that proved installation, integrations, store workflows, and production support in live service before broader expansion.

    WithOperations, support, Dunkin' teams

    05Results

    The license plate recognition experience went live across 50 Dunkin' locations and created demand for broader expansion, plus interest from other clients.

    CURO Financial TechnologiesConsumer lending

    Turning lending complexity into clear customer journeys.

    A lending journey has to feel simple to the customer while staying correct across lending rules, transactions, integrations, and store operations. At CURO, I owned product initiatives for Cash Money's Canadian lending experience, online and in stores.

    Omnichannel scope: web and storefront lending journeys in Canada
    Employer
    CURO Financial Technologies
    Role
    Product Manager
    Product
    Cash Money, Canadian consumer lending
    Surfaces
    Desktop and mobile web, storefront systems
    I owned

    Requirements, user stories, acceptance criteria, and a shared, prioritized backlog for the lending journeys.

    With the team

    Alignment with Engineering, Data, Legal, Marketing, Product Marketing, and storefront operations. Engineering built the journeys; I coached newly transitioned Product Owners.

    Outcome

    Clearer requirements for forms, navigation, and transaction-critical steps, and a stronger product practice during the Agile transition.

    Key product decisions
    • Rules into requirements. Translated lending rules and system dependencies into testable requirements and acceptance criteria.
    • One journey, online and in store. Aligned the web and storefront journey so the store saw the same loan the customer saw online.
    • Legal and operations early. Brought Legal and operations into transaction-critical steps before build.
    Customer journey mappingProduct requirementsAcceptance criteriaBacklog prioritizationStakeholder alignmentRegulated workflows
    One loan screen, x-rayedA sample screen I made, not the real Cash Money app

    Explore a customer’s loan review screen. Select a field to see the lending rules, data, and systems behind it, and how that information connects to the in-store experience.

    Field 1 of 6
    cashmoney.ca/apply
    Step 3 of 4Review your loan
    Systems behind this fieldSimulated
    Lending rules
    Customer data
    Compliance
    Banking
    Store system
    Loan amount

      Before I wrote a single requirement, I mapped every field like this. It's how I learned that one "simple" field on a loan screen is usually three teams' problem.

      In-store view · Ontario branchPOS-02
      CustomerM. Thompson · online application
      Loan$500.00 · due next payday
      StatusIn progress · step 3 of 4
      Read the full storyHide the full story

      01Customer journeys

      Customers needed loan steps they could understand and finish, online or in a store. With Engineering and business stakeholders, I mapped each customer action to its lending rule, data dependency, and operational consequence before defining any user-facing behavior.

      WithEngineering, Data, business stakeholders

      02Requirements

      I turned that map into requirements, user stories, and acceptance criteria that made dependencies and edge cases testable instead of implicit.

      WithEngineering, Data

      03Backlog priorities

      I kept one shared, prioritized backlog, because a change to one customer step could affect several systems and teams.

      WithEngineering

      04Stakeholder alignment

      Transaction-critical steps got Legal and operations input early. Requirements were validated with Engineering against the business logic and reviewed with Legal, Data, and storefront operations before build, so the store saw the same loan the customer saw online.

      WithLegal, Data, Marketing, Product Marketing, storefront operations, executives

      05Delivery practices

      During the move to Agile, I helped newly transitioned Product Owners develop backlogs, stories, acceptance criteria, and working practices.

      What I took from it: in a regulated product, the fastest path to a simpler screen is getting Legal and operations in the room before the design exists.

      WithProduct Owners

      Purple Peacock StudiosFounder chapter

      Customer experience and commercial ownership.

      I co-founded Purple Peacock Studios and built the customer journey around the creative service, from the first inquiry through consultation, booking, the wedding day, editing, and delivery.

      300+ weddings and events · 200+ inquiries a year · multiple U.S. markets
      Company
      Purple Peacock Studios (co-founded)
      Role
      Co-Founder | Product, Growth
      Product
      Wedding and event photography and film
      Surfaces
      Website, inquiry and booking workflow, in-person experience
      I owned

      Positioning, pricing, packaging, acquisition, inquiry-to-booking decisions, and delivery workflows.

      With the team

      Service decisions with my co-founder; workflows with the contractors, editors, and vendors who deliver the shoots and galleries.

      Key product decisions
      • Clear packages, real capacity. Kept packages easy to understand and trustworthy, without promising more than the team can deliver.
      • Discovery to booking. Connected SEO and acquisition to inquiry tracking, consultation, and booking.
      • Repeatable delivery. Designed repeatable production and post-production workflows, with AI-assisted culling and editing, while protecting creative quality.
      Customer discoveryPricing & packagingGo-to-marketConversion thinkingWorkflow designBusiness ownership
      Explore the customer journeyA sample couple I made up, not a real client

      Follow a couple from discovering Purple Peacock to receiving their photos and film. See their experience alongside the decisions and workflows behind it.

      Discovery
      What the couple sees
      Studio pipelineSimulated
      Sarah & JamesOct 12 · Chicago · Photo + filmNew inquiry
      SEO & marketing
      Inquiry tracking
      Consultation flow
      Capacity check
      Crew & venues
      AI culling
      Client feedback
      Studio log
        Read the full storyHide the full story

        01Customer discovery

        A couple judges style, trust, price, and availability long before the wedding, and remembers delivery long after it. Customer conversations, inquiry patterns, and market feedback show me what clients value, where decisions feel hard, and what reassurance they need before booking.

        WithClients, my co-founder

        02Positioning

        I own positioning, and I use inquiry patterns and conversion signals to check whether our messaging connects with the couples we want to serve.

        WithMy co-founder

        03Pricing & packaging

        Packages need to stay easy to understand while flexing for different events, needs, and levels of service. Pricing and capacity shape each other: the goal is a clear consultation and booking process that never promises more than the team can deliver.

        WithMy co-founder

        04Acquisition

        I own acquisition decisions, including SEO and marketing, and track each inquiry from first contact through consultation and booking.

        05Service design & operations

        I design repeatable workflows across production and post-production, standardizing handoffs and bringing in automation and AI-assisted culling and editing, while protecting creative quality and the client experience.

        WithContractors, editors, vendors

        06Business outcomes

        This is where my ownership is most direct. Pricing, conversion, capacity, and quality all show up in the same customer's experience, and I see the consequences in bookings, delivery, and client expectations.

        Builds

        Things I've built on my own.

        Building is how I stay close to how AI products really behave. Each one says plainly whether it's a public demo, a prototype, or still in progress.

        CloutRentals + Clout Concierge · Full-stack AI product · Exotic car rentals, Chicago

        From customer questions to actionable leads.

        Public demo · Production launch coming soon

        People renting a supercar ask the same questions about availability, pricing, and rules at any hour, and every wrong answer about a policy is a real business risk for the owner. I designed and built a rental website and AI concierge that answers from approved information and turns questions into leads the owner can act on.

        Built for
        CloutRentals, an exotic car rental business in Chicago
        Role
        Product owner, designer, and builder
        Designed in
        Claude Design, then built and deployed by me
        I owned

        Everything: requirements, design, front end, back end, admin panel, and deployment.

        With the owner

        Which policies the AI may use. He follows up on leads from the admin panel.

        Status

        Public demo: site, concierge, lead capture, email notifications, and admin are built and working. Production launch coming soon. Planned measurement: answer quality, lead completeness, and inquiry-to-booking conversion.

        Key product decisions
        • One workflow first. Scoped the first version around one complete inquiry-to-follow-up workflow.
        • Grounded answers. The AI answers only from owner-approved information and says when it doesn't know.
        • Actionable leads. Each inquiry becomes a lead with a summary and status, and the owner gets an email.
        ReactREST API on RenderOpenAI APIMongoDBResendGitHubAI Product ManagementLLM guardrailsAPI integrationData modelingInternal tools
        Explore the inquiry-to-lead workflowSample conversation and lead, not real customer data

        A visitor asks the concierge about a car. You see the website on the left and what I built behind it on the right, from the grounded AI reply to the lead and the owner's email. It plays on its own. Pause anytime, or click a step.

        7:14 PM
        cloutrentals.com
        Drive something unforgettable.Premium supercars · Chicago
        Clout ConciergeUsually replies instantly
        Type a message…
        What runs behind itSimulated
        Chat widget
        REST API · Render
        OpenAI
        MongoDB
        Resend
        Admin panel
        Admin panelLeadsPoliciesSettings
        Email sent · New lead: Alex wants the Lamborghini on Saturday
        Alex RiveraLamborghini · Saturday · birthday
        new
        Cancellation policyapproved · AI may use
        Out-of-state travelplaceholder · AI says unknown
        API log
          Read the full storyHide the full story

          01First-version scope

          The owner's real bottleneck was turning scattered questions into leads he could act on. I scoped the first version around one complete inquiry workflow: answer questions from approved information, capture lead details, store inquiries, notify the owner, and support follow-up through an admin interface.

          WithThe owner

          02AI guardrails

          The AI only uses policies the owner has approved. For anything still marked as a placeholder, it says it doesn't have verified information and passes the question on. In a business with deposits and insurance on the line, a confident wrong answer is worse than no answer.

          WithThe owner

          03Lead workflow

          • Two ways in, one pipeline: chat or the reservation form both land in the same admin.
          • Leads you can act on: each lead gets a summary, vehicle, date, contact details, and a status (new, contacted, booked, closed), with the full transcript one click away.
          • Owner controls without code: an on/off switch for the concierge, reply time shown to customers, emergency contact, and where notifications go.

          04Deployment

          It runs as a public demo, with a production launch coming soon: a React front end with fleet pages, a reservation form, FAQ, and the concierge widget; a REST API on Render; OpenAI for replies; MongoDB for leads, reservations, conversations, policies, and settings; and Resend for email. When a lead comes in, the owner gets an email and follows up personally from a key-protected admin panel.

          05Planned measurement

          The plan is to use transcripts as the eval set: each “I don't know” points to a policy the owner still needs to confirm, and approving it improves the AI without touching a prompt. Not measured yet; next up: answer quality, lead completeness, owner follow-up, inquiry-to-booking conversion by channel, and whether a live human handoff is worth adding.

          What I took from it: the model is rarely the hard part. The hard part is deciding what it's allowed to say, what happens when it doesn't know, and who hears about it.

          Mobile · Audio · Music education

          SurAI

          In development · V1

          A vocal training companion for singers practicing Hindustani and Western music. It listens through the microphone, detects pitch, compares it with a target note or melody relative to the singer's chosen tonic, and gives immediate feedback in cents.

          MVP decision

          V1 tests one practice loop: pick one of two songs, sing a phrase, see live pitch feedback, and try again. Western notes by default, Sargam one tap away. A song library, ornament lessons (meend, murki, gamak), streaks, and accounts wait until the loop is proven.

          User hypothesis

          Does hearing real-time feedback make someone want to try the same phrase again? Success looks like people tapping Try Again and improving across attempts. I'd also test how strict the "on pitch" window should be.

          I'm building this solo to explore how immediate feedback can make music practice more useful and approachable.

          Explore pitch feedbackSimulated voice, not a real recording

          Someone sings a five-note phrase against faint target bars, and SurAI shows in cents whether they're sharp, flat, or on pitch. It plays on its own. Drag the slider to sing it yourself.

          Target
          DoC4
          sung ±0¢ · 261.6 Hzon pitch
          Nice and steady.

          The hard product question

          How much deviation should count as "on pitch"? Too strict and beginners quit. Too loose and the feedback is useless. The demo uses a ±10 cent window as a test assumption, not a validated threshold.

          See the V1 wireframeHide the V1 wireframe

          Low fidelity on purpose: structure, flow, and what is left out of V1. All numbers on screen are sample data, not test results.

          Problem (working assumption)People who want to sing better often practice alone and can't tell whether they are on pitch. A video can't listen to them, and a teacher isn't always available.
          Who it's forSelf-taught singers who practice at home and want to match real songs they already love.
          What V1 testsDoes hearing real-time feedback make someone want to try the same phrase again?
          What success looks likePeople finish a take, tap Try Again, and their accuracy improves across attempts.
          1 Choose notation→2 Pick a song→3 Practice→4 Singing→5 Result→6 ProgressTry Again loops 5 → 37 to 9: mic or signal fails
          3 · Practice ready
          ‹ BackPhrase 1Key: Original
          Someone Like You
          Adele
          C D ESa Re Ga
          Opening line of verse 1
          Start on the first note
          note 1 · note 2 · note 3 · note 4 · note 5
          target melody
          Slow
          Sing
          Hear
          Try AgainCompare
          DecisionThe pitch stage is the largest element. Try Again and Compare stay muted until there is a take, so the first action is obvious.
          5 · Result sample data
          ‹ BackPhrase 1Key: Original
          Opening line of verse 1
          78% accuracy. Running sharp, about 15¢ high on average.
          target · your voice
          Try Again
          Hear target
          View progressNext phrase (not in V1)
          DecisionOne number and one plain sentence. Correction should feel constructive, not like a grade. Try Again is the primary action.
          8 · No clear note weak signal
          ‹ BackPhrase 1
          Opening line of verse 1
          We couldn't pick up a clear note. Try moving closer to the mic or singing a bit louder.
          Try Again
          Hear target
          DecisionNo score when the signal is weak. A wrong score would be worse than none, because it teaches the singer the wrong thing.

          Biggest risk: trust in the feedback. If the app says a note is flat when it isn't, people stop believing it, and the whole product depends on that belief.

          How V1 handles it: it never shows a confident score when the signal is weak, and says so in plain words. How to test it: sing known notes and compare the app's reading against a tuner.

          Open the full wireframe: all 9 screens and what's cut from V1 ↗

          Visual directionA concept for where Sur could go after V1, not the V1 scope. © 2026 Yamna Irfan. All rights reserved. Shared for portfolio review only; not licensed for reuse.
          Sur visual concept: brand palette, practice screen, and song list
          Conversational AI · LLM

          Thought Untangler

          Prototype

          User problem: an anxious thought is hard to untangle alone, and a general chatbot tends to jump straight to advice.

          Interaction decision: one guided flow that walks someone through the CBT downward-arrow technique, from a fearful thought to the belief underneath it, with Socratic questions instead of advice.

          TUThought Untangler

          Validation plan: before expanding, test whether people find the questions helpful and safe, with input from mental health professionals. It's an LLM prototype, not a clinical tool.

          Built with: React, plus structured prompts and system instructions that keep the model asking instead of advising, and tone guardrails that keep it warm without playing therapist.

          Interaction design

          Somatic Body-Mapping

          Prototype

          User problem: free-text journaling about a feeling takes effort, especially in the moment.

          Interaction decision: a tap-based body map for logging where a feeling shows up, region by region, so input is faster and lower-effort but still produces a useful record.

          Tap where you feel it0 logged
          Chest What does it feel like?
          Intensity 5/10
          Save

          Validation plan: test whether tapping is faster and easier than free-text journaling while still producing a record people find useful.

          Why I build these: they're fast ways to test conversational AI and low-effort interaction patterns with real people, the same questions behind any AI-powered customer experience.

          How I work

          From people to launch, and back again.

          How I take a product from understanding people to launch and learning, the teams I bring in along the way, and how I think. Hover or click anything in the diagram to trace it back to me.

            The customer question

            The product decision

            Who I work with

            Where you can see it

            1. 01Customer discoveryNeeds, behaviors, and workflows
            2. 02Product strategy & success measuresThe problem, the goal, and the KPIs
            3. 03MVP scoping & validationThe smallest useful version, tested early
            4. 04Requirements & collaborative deliveryStories, criteria, and backlog, with Engineering and Design
            5. 05Go-to-market & phased rolloutReadiness, UAT, and rollout in phases
            6. 06Experimentation & iterationAdoption, engagement, and what to fix next

            How I talk to them
            Skills

            What I bring to the table.

            Grouped by the work I do. Each group links to where you can see it in practice.

            Customer discovery & experience

            • Customer discovery & user research
            • Customer journey mapping
            • Consumer psychology
            • Human-centered design

            See it at Meta · Purple Peacock · CURO

            Strategy & execution

            • Product strategy & roadmapping
            • MVP scoping
            • PRDs, user stories & acceptance criteria
            • Backlog prioritization
            • UAT & release management
            • Stakeholder alignment & cross-functional collaboration

            See it at Coates · CURO · CloutRentals

            Launch & learning

            • Go-to-market & phased rollout
            • Pricing & packaging
            • Experimentation & product analytics
            • KPI definition & tracking
            • Adoption & engagement

            See it at Meta · Purple Peacock · Coates

            Technical product fluency

            • Applied AI
            • Voice & conversational AI
            • AI evaluation
            • Systems & hardware/software integration

            See it at Meta · Coates · CloutRentals

            Product & delivery

            Jira · Confluence · Asana · Figma

            AI

            Claude · Claude Code · Claude Design · ChatGPT · Gemini · OpenAI API

            Build

            Emergent · Base44 · React · REST APIs · MongoDB · Render · Resend · GitHub

            Contact

            Say hi.

            I’m looking for my next Product Manager or Senior Product Manager role, bringing customer discovery, technical fluency, and ownership from early ideas through launch and iteration.

            I’m especially interested in AI-powered customer experiences and connected systems.

            Chicago-based · Open to remote, hybrid, onsite, and relocation

            Have a role in mind? I’d love to hear about it.