Sergey Yaremenko
Intelligence Product Builder

I build intelligence systems that turn data into decisions.

15 years across product, data, ML, and AI. Knowledge graphs, multi-agent search, anomaly detection, decision-support layers.

▸Knowledge graphs▸Multi-agent systems▸Anomaly detection▸Scoring & matching▸Predictive models▸RAG pipelines▸Hybrid search▸CUPED▸Uplift modeling▸Sequential testing▸ClickHouse▸Neo4j▸BigQuery▸Snowflake▸Redshift▸dbt▸Prefect▸Airflow▸Pydantic AI▸Vertex AI▸Claude▸Gemini▸Python▸SQL▸R▸TypeScript▸Supabase▸Vercel▸Knowledge graphs▸Multi-agent systems▸Anomaly detection▸Scoring & matching▸Predictive models▸RAG pipelines▸Hybrid search▸CUPED▸Uplift modeling▸Sequential testing▸ClickHouse▸Neo4j▸BigQuery▸Snowflake▸Redshift▸dbt▸Prefect▸Airflow▸Pydantic AI▸Vertex AI▸Claude▸Gemini▸Python▸SQL▸R▸TypeScript▸Supabase▸Vercel
01What I build

Cases: click to open details

  1. case · 01open →
    DOGER
    doger.ai
    regulatory filingscorporate registrymarket datatrade flowssentimentNEO4Jknowledge graphsearchanomalyranking
    open-data graph stack

    Sovereign decision intelligence for European public data — a self-expanding agentic decision-support system.

    One natural-language question fans out to domain agents in parallel and returns a single cited answer. Site selection, wildfire exposure, loan risk, cross-border ownership — joined across public datasets nobody publishes together.

    self-expanding graph · EU-sovereign · cited answers in seconds
    • Neo4j
    • Pydantic AI
    • Prefect
    • XGBoost
    • Claude
    case studyfull case →
  2. case · 02open →
    Scaleo
    scaleo.io
    F140%F225%F320%F415%ANOMALY FEEDS214:32spike +218%S113:08drop -64%S311:21rate-of-changeS209:55deviation > 3σ
    match score · live alerts

    Built a new AI product from scratch for a partner-marketing platform

    On top of Scaleo's affiliate-tracking platform: match the right partners automatically, catch unusual activity and fraud, and predict revenue and churn.

    partner matching · predictions · fraud flags
    • ClickHouse
    • Airflow
    • Looker
    • AWS
    • LLM agents
    case studyfull case →
  3. case · 03open →
    Personalization at marketplace scale
    TRIGGERlifecycle_stage = renewalcompatibility < 0.65context_signal matchCOHORT SPLIT47%LOGGED-INlogged-in 47%anonymous 53%
    trigger · cohort · activation

    Asset-aware activation on a 9-figure-revenue marketplace

    Lifecycle-stage prediction, compatibility-aware fallbacks, context-cued recommendations on a marketplace where most buyers don't log in. Cohort and session-level scoring as the default.

    cohort + session scoring · lifecycle activation
    • Braze
    • Recommendations ML
    • BigQuery
    • Productboard
    • A/B
    anonymizedfull case →
  4. case · 04open →
    PeopleForce
    peopleforce.io
    $787K · Q3$1.8M · Q2+128% ARR
    ARR · 4-quarter view

    Took a $787K HRM SaaS through a wartime pivot to $1.8M

    Inherited zero roadmap, zero repeatable discovery, unpredictable delivery. Rebuilt all three. Then went international.

    $787K → $1.8M during tenure
    • Productboard
    • Roadmap
    • Discovery
    • Cross-functional
    case studyfull case →
  5. case · 05open →
    PandaDoc
    pandadoc.com
    DOC OPEN100%100%+14ppFIRST FIELD78%92%+25ppSIGNATURE42%67%+23ppCOMPLETE18%41%beforeafter
    completion funnel · before vs after

    Cut friction in the editor that gated expansion revenue

    Owned Editor 2.0. Time-to-complete down, satisfaction up. The discovery framework I introduced became the team's standard.

    $35M → $55M revenue era · $1B unicorn
    • Discovery
    • Product analytics
    • A/B testing
    • User research
    case studyfull case →
  6. case · 06open →
    ЛУН
    lun.ua
    SLOTBEFOREAFTERLIFTP1$2.10$2.78+32%P2$1.45$1.92+33%P3$0.95$1.20+27%P4$0.62$0.74+19%
    rev / impression by slot

    Redesigned ad ranking on the dominant real-estate marketplace

    ML ranking + analytical tooling on both sides of the marketplace. Shipped to 50+ enterprise clients without breaking engagement.

    +15–30% revenue / customer · 40% faster ops
    • ML ranking
    • ClickHouse
    • Python
    • SQL
    • BI
    case studyfull case →
  7. case · 07open →
    MacPaw
    macpaw.com
    snowplowS3glueredshiftlookerDATAHUB · TB / MO · 5 PRODUCTSTEAM1 → 10
    DataHub · ETL pipeline

    Built analytics from zero — Setapp launch underwritten by it

    DataHub on AWS + Redshift. The MarTech function from one analyst to a team of ten. Monte Carlo unit economics behind the Setapp launch.

    1 → 10 team · TB / month · $23M revenue era
    • Redshift
    • AWS
    • ETL
    • Monte Carlo
    • BI
    case studyfull case →
0107
↓ Get in touch
03Contact / engagements

Let's build the intelligence layer your data is missing.

Open to fractional/advisory engagements, mentorship windows, and workshops. Founders, CPOs, data leaders building intelligence-shaped products.

Book a 30-minute intro
Email
npkrol@gmail.com
LinkedIn
linkedin.com/in/npkrol
Telegram
t.me/npkrol