A career built at the intersection of models, markets and people.

Product Marketing MBA InternJun to Aug 2026

Built a translation-validation proof of concept, shaped a six-month roadmap with broad stakeholder input, and developed launch messaging for AI sales enablement.

• 50StakeholdersAI Sales Enablement AgentRoadmap, stakeholder alignment and personas for a sales-facing AI agent6 mo.Roadmap

  A six-month roadmap and persona-driven UX flow for an AI sales enablement agent, built on a 50-person stakeholder alignment process.

  5+ Personas defined

  01How I approached itDefined a six-month roadmap and interfaced with engineering on technical feasibility, which changed the design flow and end-user interaction.

  • Ran a stakeholder alignment process across 50 people spanning sales, localization and content teams to pressure-test what "good enough" looked like before proposing the roadmap.
  • Defined 5+ core personas from that research, and defined a post-research UX flow that shipped as designed.

  02Impact to usersLaunch materials and positioning from this work went live as part of Cisco's global CMO keynote on August 24.

• 15%Token reductionTranslation ValidatorTranslation-validation POC catching meaning drift before content ships

  A proof of concept that checks translated sales content against its source meaning before it ships, used directly by sales teams.

  01How I approached itBuilt using Cursor and Claude Code as a proof of concept that validates translations without re-running full inference on every piece of content.

  02How the product worksChecks translated sales content against source meaning at the point of publish, flagging drift for human review rather than blocking output.

  • Sits inside the existing content pipeline rather than replacing it.
  • Used directly by sales teams as an end product, not just an internal engineering utility.

  03Impact to usersGave sales teams a credible way to trust AI-translated content at scale.

  • Components were reused across four internal tools: content evaluation, research, mail, and PowerPoint.

Product ManagerFeb 2025 to Aug 2025

Owned two ML-backed product bets: a new in-app advertising surface and an internal platform that made model workflows easier for operators to use.

• $8MRevenue · first 60 daysIn-App Advertising LaunchNew ML-backed monetization surface$20MRevenue · 1-year mark

  A new in-app advertising surface built as a fresh monetization line, led across a 10-person cross-functional team.

  5% Retention lift

  01How I approached itLed a 10-person cross-functional team and drove the vendor partnerships required for launch.

  • Negotiated pricing-model and impression-counting definitions directly with Legal.
  • Partnered directly with Data Science, who set the quality benchmarks; I pressure-tested and signed off on those benchmarks.
  • 250M-user segmentation and cohort experiments fed targeting and placement testing for this same launch.

  02How the product worksML-backed ad targeting matched ads to user context rather than static inventory, operating on a 250M+ user platform.

  03Impact to users$8M in revenue in the first 60 days.

  • $20M in revenue at the 1-year mark.
  • 5% retention lift attributed to the launch, the only platform change introduced at the time.

• 100+Daily active users · from 8ML Application PlatformInternal platform for operator-facing model workflows$500KAnnual savings

  A separate internal platform that made ML model workflows usable for operators, scaling from 8 to 100+ daily active users.

  01The problemOperators needed to work with ML models without ML expertise.

  • Manual, ad hoc workflows were costing time and money at scale.

  02How I approached itRedesigned the developer experience for operators.

  • Partnered with Data Science on quality benchmarks.
  • Cut third-party reliance, part of how the platform generated its $500K in annual savings, not just a headcount or tooling-cost story.

  03Impact to usersScaled from 8 to 100+ daily active users.

  • Cut third-party reliance.
  • Saved $500K annually.

Product & GTMAug 2024 to Jan 2025

Used buyer research and go-to-market strategy to shape adoption for an LLM platform serving government and enterprise contexts.

• $700KNew ARRLLM Platform Go-to-MarketBuyer research and GTM for a government/enterprise LLM platformB2BBuyer research

  Buyer and market research that clarified high-value use cases and go-to-market direction for an LLM platform serving government and enterprise buyers.

  01The problemGovernment and enterprise buyers needed a clear case for adopting an LLM platform inside real institutional constraints.

  02How I approached itA GTM and research role, not build ownership: ran buyer and market research to surface which use cases mattered to government and enterprise buyers.

  • The $700K ARR client came through a formal sales process: demos, discovery calls, pitching decision-makers.

  03Impact to users$700K in new B2B ARR.

• 0 → 200KTeachers · 2 monthsStudent Assessment App LaunchAdoption-data-driven GTM for a student-testing app

  A GTM rollout for a student-testing app, driven by adoption-data analysis of lagging districts rather than an assumed rollout plan.

  01The problemAdoption data showed specific districts lagging on a student-testing app, identified through my own analysis rather than assumed.

  02How I approached itAnalyzed adoption data via Tableau to identify the lagging districts.

  • Tailored the app's go-to-market strategy and worked directly with State and Education department stakeholders.

  03Tools and product decisionsTableau: adoption-data analysis.

  04Impact to usersGrew from zero to 200K teachers within 2 months.

Associate Product ManagerJul 2023 to Jul 2024

Turned experimentation and traffic intelligence into repeatable product systems for advertising teams.

• 30Experiments · 6 monthsFeature-Flag Experimentation PlatformOrg-wide A/B testing platform with role-based accessRBACAccess control

  An org-wide feature-flag A/B testing platform with role-scoped access, built around role-based access as the core design decision.

  01How the product worksAn org-wide A/B testing platform with role-scoped access and permissions.

  • Reduced the time to start and monitor experiments, and ensured mutual exclusivity by design: experiments don't collide.
  • RBAC was introduced specifically to govern sensitive data on the platform.
  • Within 6 months of launch it was running 30 experiments.

  02Tools and product decisionsRole-based access was the core design decision: the platform's actual differentiator, not just a nice-to-have.

• $35MRevenueB2B Ad Traffic OptimizerTraffic-shaping system generating $35M revenue at 70% lower cost70%Lower cost

  A traffic-shaping system that reallocates ad traffic-acquisition spend toward higher-yield sources.

  01How the product worksShapes and reallocates traffic-acquisition spend based on yield signals, lowering cost per unit of revenue generated.

  02Impact to users$35M in revenue at 70% lower cost than the prior approach.

• $40MRevenueAPAC Go-to-Market LaunchCompetitive analysis and launch strategy for APAC expansion11New customers

  A competitive analysis and go-to-market plan that secured executive buy-in for Media.net's APAC expansion.

  01How I approached itBuilt a competitive analysis covering 5 competitors: 3 similar-size, 2 larger players.

  • This was a core standalone deliverable that secured executive buy-in for the launch.
  • Followed by a detailed channel strategy, partnership strategy, and go-to-market plan.

  02Impact to users$40M in revenue across 11 new customers within 6 months of launch.

Software Development EngineerJul 2021 to Jul 2023

Built the ML and data foundation that later informed how I lead products.

• 5TB/dayAd-transaction data processedAd Pricing OptimizerBid-prediction ML model for real-time margin management

  A bid-prediction ML model that sets margins dynamically inside the real-time ad-exchange bidding process, on the demand side.

  01How I approached itBuilt the underlying bid-prediction ML model that powers dynamic margin management.

  02How the product worksThe bid-prediction model sets margins dynamically as part of the demand-side, real-time ad-exchange bidding process, rather than using fixed or manually-set margins.

  03Tools and product decisionsBuilt in Python, processing 5TB of ad-transaction data daily, extracted via Spark SQL from HDFS.

• +20%Return on ad spendAI User Targeting ModuleBusiness case, spec and 3-person build for an ad-targeting module10th → 3rdClient-report ranking

  An ad-targeting module I wrote the business case and technical spec for, then led a 3-person engineering team to ship.

  01How I approached itWrote the business case and technical spec.

  • Led a 3-person engineering team through the build.

  02Impact to usersIncreased return on ad spend by 20%.

  • Moved Media.net's client-report ranking from 10th to 3rd.

ML EngineerJun 2019 to Jul 2021

Applied computer vision and color science to flagship mobile experiences at global scale.

• $1BDevice-line sales · contributedNight ModeLow-light computer vision for a flagship camera line

  Low-light performance work for a flagship camera line, led through beta user research and competitor analysis.

  01The problemA flagship camera line needed low-light performance that held up at mass-market scale, not just in a lab.

  02How I approached itLed development.

  • Conducted beta user research.
  • Conducted competitor analysis.

  03Impact to usersShipped as a flagship differentiator.

  • Contributed to $1B in device-line sales.

• 25M+Flagship devicesColor-Tuning ModuleML-based color tuning across a global device line

  An ML-based color-tuning module built with Korean and Vietnamese teams, optimizing on-device latency without compromising image quality.

  01How I approached itCollaborated with Korean and Vietnamese teams.

  • Built an ML-based color-tuning module.
  • Optimized on-device latency without compromising image quality.

  02Impact to usersLaunched globally on 25M+ flagship devices.

• 1 of 20Engineers selectedSuper Slow-Motion FeatureCross-border co-development of a headline camera feature2020 →Standard feature since

  Samsung Camera's first software-based Super Slow-motion feature, co-developed with the Korea R&D Center as 1 of 20 selected engineers.

  01How I approached itSelected as 1 of 20 engineers.

  • Co-developed the feature with the Korea R&D Center.
  • Worked cross-border on a headline camera feature.

  02Impact to usersShipped as Samsung Camera's first software-based Super Slow-motion feature.

  • Has remained a standard feature across devices since 2020.