Skip to main content
Liorry Herisnor

Case Studies

Every project below states its current status and what I personally owned. The main case studies carry the full write-up: the business problem, how the work used to be done, goals and non-goals, my role and who supported delivery, the solution, where AI is used and where it is deliberately not, testing, results and lessons.

What the status labels mean

Status: Production
Running in production and carrying real work today.
Status: Shipped
Delivered and used. Not under active development.
Status: Live internal tool
In daily use internally rather than shipped to users.
Status: Partially operational
Some parts run in production; others are built but not yet wired in. The case study says which.
Filter by category
Filter by status

Showing 10 of 10 projects

All projects

Sorted by flagship work first, then most recently updated, then supporting builds and roadmap items.

  • Illustration of the lead discovery and enrichment architecture: a job definition feeds seeded query expansion, a query cache and metasearch with fallback; then fetch and extract, validate and classify, an operator review gate, and a sending list; then an enrichment queue, a skip-locked worker, heartbeat and reaper monitoring, and a segmentation record.
    Status: ShippedData and Analytics

    Lead Discovery and Enrichment

    A discovery and enrichment pipeline that finds publicly published business contacts and enriches records for segmentation — with a mandatory human review step before anything enters a sending list.

    Discovery volume
    300K+ contacts / month
    • Go
    • PostgreSQL
    • Redis
    • +3 more

    2026 — present · Product owner, pipeline architect and delivery lead

  • AdSpy search overview: header counts of 66,870 ads and 114,098 landing pages, per-network totals for Dianomi, Facebook, MGID, Outbrain, Revcontent and Taboola, a filter bar with winner score, expected CTR, advertiser, geo, device and keyword fields, a facet row of networks, countries, devices and tracker signatures, and the first row of captured ad creatives. Account identifier covered.
    Status: ProductionMarketing and Growth

    AdSpy — Competitive Ad Intelligence

    A pipeline that captures native ads and their landing pages at scale, deduplicates them visually, resolves advertiser identities, and connects the findings directly to campaign creation.

    Corpus
    66,870 ads · 114,098 pages
    • Go
    • PostgreSQL
    • Perceptual hashing
    • +2 more

    2026 — present · Product owner, architect and delivery lead

  • Illustration of the AdFetch architecture: a residential proxy pool, browser workers, a job queue and a local scrape endpoint form the fetch and scrape core; a native fetch API and a Firecrawl-shaped scrape, search and map surface sit alongside upstream Firecrawl and a rule against ever holding ad or lead logic; a per-consumer byte budget and a hard stop on overage govern two named clients, AdSpy and Lead Discovery.
    Status: ProductionInfrastructure

    AdFetch — Shared Fetch, Scrape and Search Infrastructure

    A self-hosted, Firecrawl-compatible fetch, scrape, search and map service. Headless, deliberately generic, and called independently by two other products in this portfolio.

    Consumers
    2 independent products
    • Go
    • PostgreSQL
    • Redis Streams
    • +4 more

    2026 — present · Product owner, architect and delivery lead

  • Illustration of the Elite Modern Marketing operating system: an inbound inquiry is qualified and acknowledged by AI, routed and scored by fixed rules, then handed to a person before any commitment; a discovery conversation and AI brief synthesis produce an approved structured brief; AI content and creative production is then edited, selected and finished by a human, passes a quality review gate, and feeds performance data back.
    Status: ProductionMarketing and Growth

    AI-Powered Business Operations Platform

    The platform behind Elite Modern Marketing — AI-assisted client intake, content and creative production, and reporting, with a person accountable for every client commitment.

    Clients served
    170+
    • Large language models
    • Generative image tooling
    • Workflow automation
    • +2 more

    2023 — present · Founder, product owner and operations lead

  • No screenshots. These were client production environments; nothing from them is publishable.

    Status: ShippedEnterprise Experience

    Enterprise Technology Programs in Financial Services

    Four consulting engagements inside large financial institutions — requirements, structured testing, defect management and release readiness across Loan IQ, a $45m program and advisor analytics.

    Program scale
    $45m budget
    • Loan IQ
    • SQL
    • SAS
    • +5 more

    2011 — 2015 · Business analyst and delivery consultant

  • Illustration of the AI-assisted engineering workflow: a written requirement passes sensitivity classification, context-pack assembly and routing by task class; then AI implementation against a task card, human review, automated commit checks with no AI involved, and AI pre-push review; then human deploy approval, a flag defaulting to off, post-deploy verification, and an invariant or runbook as the output.
    Status: Live internal toolProduct Delivery

    Using AI to Improve Product Development

    The development process behind the other case studies — written requirements, scoped tasks, AI coding tools, automated pre-release checks and human approval on every production change.

    Development cycles
    50% faster
    • Claude and Claude Code
    • Repository operating manuals
    • Context packs and task cards
    • +3 more

    2025 — present · Author and owner of the workflow

  • The platform's operations console showing global status, a platform-to-engine handoff readiness checklist with pending and completed steps, per-engine state and a live event log
    Status: ProductionPlatforms

    Modernizing a Legacy Marketing Platform

    The program view of a marketing operation rebuilt as a product — one compliance check every send must pass, automatic delivery protection, hands-off recovery, and releases that ship switched off.

    Compliance enforcement
    One mandatory gate
    • Go
    • Rust
    • PostgreSQL
    • +4 more

    2025 — present · Product owner and program lead

  • MailerClub campaign detail view showing the delivery funnel from injected through delivered, with per-stage bounce, deferral and complaint counters
    Status: ProductionPlatforms

    MailerClub — Email Marketing Platform

    A Go platform that takes a campaign from recipient list to delivered mail — unsubscribes and bounces enforced on every send, automatic delivery protection, and full event reporting.

    Core loop
    Suppression-first
    • Go
    • PostgreSQL
    • Redis
    • +5 more

    2025 — present · Product owner, systems architect and delivery lead

  • giMTA fleet view: six lanes all healthy with 48 engines live, lane identity and address columns blurred
    Status: ProductionInfrastructure

    giMTA — High-Volume Email Delivery Engine

    A purpose-built Rust mail server for high-volume sending — it resumes after an interruption without re-sending a single message, and learns each provider's rate limits instead of guessing them.

    Recovery model
    Zero double-sends
    • Rust
    • tokio
    • SMTP
    • +4 more

    2026 — present · Product owner, architect and delivery lead

  • Illustration of the AgentOps governance architecture: a scheduled job passes through a deduplication window, a rate limit and a capability registry, then policy mode resolution, leader election, hybrid retrieval and bounded reasoning, producing an attempt record, an operator disposition, escalation evidence, and a fail-closed-on-error rule.
    Status: Partially operationalAI Agents

    AI Agent Governance and Knowledge Platform

    The controls that let AI agents work against a live system — spending limits, an explicit list of permitted actions, a staged rollout, and search that grounds them in real documentation.

    Rollout model
    Observe → assist → autofix
    • Go
    • PostgreSQL
    • Redis
    • +4 more

    2026 — present · Product owner, governance architect and delivery lead