DK">
Work AI Lab Side Project About Contact Download CV
Senior Product Manager

Building 0-to-1 Products From Discovery Through Profitable Launch

Product leader with 6+ years shipping AI/ML platforms, IoT products, and GTM-ready launches. I take ideas from customer discovery through production deployment, technical validation, and go-to-market execution. Track record includes leading product on a $50M ARR platform, launching products that generated $12M new ARR, and shipping partner integrations that scaled from 10 to 50+ live integrations.

Core Strengths

AI Dashcam Platform Launch

Discovery to GA launch, cross-functional execution across firmware, cloud, ML, and GTM.

Partner Channel & Commercial Strategy

Pricing, packaging, co-sell motions, and enablement programs that scale.

ML Risk Scoring & Data Platform

Model roadmap, data products, outcomes measurement, and monetization strategy.

View My Work Get in Touch
$50M
Platform Scope
500K+
Subscriptions
6+
Years Experience
$12M new ARR (year one) 50+ partner integrations 25% TAM expansion 5 concurrent workstreams 35+ engineers aligned
Impact

Core Case Studies

Featured · 2026 · Solo end-to-end build

Fleet Safety GenAI

End-to-end gen-AI + agentic-AI system for fleet driver-safety coaching

Fleets generate enormous dashcam-event telemetry, but traditional tools dump it as lists for a manager to read, interpret, write up, send, and never measure. Built a working browser-demoable system that replaces that cognitive load with a governed agentic pipeline — identify the pattern, draft personalized coaching grounded in the driver's actual events, apply guardrails, queue for human approval, and measure the outcome against a matched control cohort.

33%Median event reduction (n=9, diff-in-diff vs control)
60+ minManager time reclaimed weekly
6Flows shipped, production-shape

A look at the system

Two surfaces on the same data: chat for ad-hoc questions, the flow builder for repeatable interventions.

Fleet Safety / Chat
Who is my most dangerous driver this week?
▍
SQL · driver_score_v · 14d
Compare distracted-driving across depots last 30 days.
▍
SQL · depot_distraction_30d
Fleet Safety / Flows / weekly-tailgating-coach Save flow

Partner Channel & Commercial Strategy

Led commercial strategy, pricing architecture, and partner enablement programs

Designed revenue-share commercial model and tiered pricing structure that scaled partner ecosystem from 10 to 50+ integrators. Built co-sell motions, created sales enablement programs (demo scripts, case studies, training materials), and established a commercial framework now adopted company-wide. Partnered with Sales and Marketing on partner onboarding, competitive positioning, and field enablement.

GTM Strategy Revenue Model Pricing Strategy Sales Enablement
50+Partners scaled
50%Faster integration
Partner Channel & Commercial Strategy - Billing System

AI Dashcam Platform Launch (AI-14)

Launch owner across firmware, cloud, ML, certification, and tooling

Led end-to-end GTM launch of next-generation AI dashcam, partnering with Sales and Marketing on positioning, competitive differentiation, and field enablement. Managed 5 concurrent workstreams (firmware, cloud, ML, carrier certification, tooling) across 35+ engineers. Structured tiered subscription pricing based on feature packages. Generated $12M ARR in year one and expanded TAM by 25% through new market segments.

GTM Launch AI/ML Computer Vision IoT
$12MNew ARR generated
25%TAM expansion
See Product Demo
Lytx AI-14 Dashcam

ML Risk Scoring & Data Platform

ML product lead translating models into shippable product surfaces

Drove ML roadmap partnering with data science to design compound risk scoring combining 7 vision models across edge and cloud. Structured risk scoring into tiered subscription packaging (basic detection, advanced analytics, predictive insights). Built the product analytics layer that links model outputs to safety outcomes and commercial KPIs, using 50M+ inference events to inform roadmap, retention motions, and upsell paths.

Machine Learning Edge + Cloud Data Platform Pricing Strategy
7Models unified
50M+Inference events analyzed
ML Vision System with ADAS Detection

Additional Leadership Work

Developer Platform & Partner Ecosystem

API product lead delivering OAuth flows, portal UX, and docs

Established API platform strategy powering AI-12/AI-14 integrations and serving 50+ partners. Defined tiered partnership model, OAuth 2.0 flows, and revenue-share framework where partners charge premium subscription costs and share revenue. Designed partner portal from wireframes to production including integrated enablement resources (demo scripts, case studies, training materials). Framework adopted company-wide.

API Design OAuth 2.0 Partner Enablement Revenue Share
50+Partner integrators
40%Reduced support tickets
Surfsight Developer Portal
API Documentation

Global Expansion & Carrier Partnerships

Drove carrier certification requirements and execution with partners

Drove global market expansion by structuring carrier partnerships as distribution channels. Negotiated commercial terms with AT&T, T-Mobile, Telstra, and Bell Mobility, defining certification framework and partnership agreements. Opened 6 new sales channels across 5 continents and expanded addressable market by 25%. Created repeatable blueprint now used across product lines.

Certification Market Expansion Carrier Partnerships Distribution Strategy
6Channels unlocked
25%TAM expansion
FCC Certification CE Certification AT&T T-Mobile Bell Mobility
Personal Project

AI Lab

Personal AI Infrastructure — CrabCake Platform

Builder & Architect — Self-hosted multi-agent platform running real workflows

Started as OpenClaw, rebuilt as CrabCake — a provider-pluggable gateway (claude-code / ollama / openai-compat) running on a Mac Mini M4 Pro. Four architectural generations in ~6 weeks: initial hub-and-spoke → Sonnet-selective rebuild (≤$1/day cost target) → ComfyUI image gen → CrabCake with streaming gateway. Every decision — which model for which task, when to run locally vs. cloud, how to manage memory and cost — directly informs how I reason about building AI products in my day job.

9Specialized agents
11Models orchestrated
≤$1Daily cloud budget
14+4Claude Code skills shipped
CrabCake Ollama · 5 local models Devstral Qwen 3.5 ComfyUI · FLUX.1 Dev Claude Sonnet 4.6 OpenAI · OpenRouter Node.js Gateway Next.js Dashboard SQLite · BM25 · RAG LLM Wiki (auto-maintained) Launchd · Cron

Hybrid Inference Orchestration

  • 11-model fleet: 5 local (Ollama — Devstral, Qwen 3.5 27B/14B/8B, custom Qwen router 4B) + 3 cloud (Sonnet 4.6, Haiku 4.5, mistral-small via OpenRouter) + ComfyUI/FLUX.1
  • ≤$1/day cloud budget — Sonnet only where quality is non-negotiable (CV pipeline); local-first by default, explicit cloud escalation
  • Serialized Claude lock + concurrent Ollama + auto-fallback when Ollama is down — no single point of failure in the request path
  • LiteLLM proxy bridges Ollama to OpenAI-compatible clients (port 4000); `ollama_chat/` prefix required for tool calls

Multi-Agent Architecture

  • 9 specialized agents (Router · Main · Career · Finance · Intel · Health · OpsSec · Writer · LLMLab) behind a pluggable provider abstraction (claude-code / ollama / openai-compat)
  • Deterministic zero-LLM regex router — <10ms dispatch, no token cost — with classifier patterns ordered so domain-specific routes win over generic matches
  • Custom Node.js gateway daemon (port 3400) with streaming responses to Telegram via edit-in-place, agent-to-agent delegation, session memory with compaction safeguards
  • VRAM-aware model loading, warm/cold state (KEEP_ALIVE=10m), FLASH_ATTENTION=1 on 48GB Mac Mini M4 Pro

Measurement & Learning

  • Next.js real-time dashboard (port 3400): VRAM, token cost, cloud vs. local spend, cron health, agent uptime
  • BM25 per-agent memory + post-trade reflection — agents learn from their own mistakes across sessions
  • Walk-forward backtesting (6-month train / 1-month validate) — no cherry-picked windows, used in the crypto trading system
  • SQLite benchmark DB + radar charts comparing model quality across coding, reasoning, writing; cron-driven reports every 2h

Security & Constraints

  • 3-layer gateway security — allowlist-based Telegram access control, request-tier classification, encrypted credential store
  • Tiered execution: automatic for low-risk ops, explicit approval gate for privileged/destructive actions (no silent writes)
  • Family Proxy via Cloudflare Tunnel — restricted reverse proxy with per-user scoping, no raw port exposure
  • Full audit trail: structured logs, credential health monitor, launchd-managed restart policies on every service
Systems Shipped From This Lab
Three products the lab actually produced — each one a standalone PM story
Research-Backed Rebuild

Adversarial Debate Trading System

Read the TradingAgents paper and replaced a flat 5-ML-engine consensus with a hierarchical debate: Bull/Bear researchers → Research Judge → Trader → Aggressive/Conservative/Neutral risk debate → Portfolio Manager final call. Documented 8 concrete failure modes before rebuilding — no speculative refactor.

27 files 2,451 LOC BM25 memory Walk-forward 6mo/1mo Quarter-Kelly sizing
Discovery Pipeline

Product-Market Radar with LLM Autotuner

Multi-topic opportunity scanner across Reddit (via Brave), Hacker News, GitHub, and Brave Search. Killed the original 7-axis fake-precision scoring and rebuilt on three honest axes — Pain (40%) · Scale (35%) · Timing (25%). Phase 3b uses a local LLM to grade outputs and refine its own queries. PM discipline applied to the PM's own tool.

4 sources 3-axis scoring LLM autotuner Cross-source multiplier Daily 02:30 cron
Meta Pipeline

Self-Tailoring CV Pipeline

JD → tailored .docx in one command. Sonnet-powered, 11 sacred base facts, 12 tailoring rules, recruiter sentence bank mined from 6 real PM CVs, a validator that rejects drift, and per-company Python generation scripts. The CV you just downloaded was produced by this system — and so were the 18 others on my desk.

11 base facts 12 rules 18 companies shipped Sonnet 4.6 required
Side Project

Tiny Wins

9:41 ⚡ 87%

Tiny Wins ⭐

Emma's Progress

⭐⭐⭐⭐⭐
🧹
Cleaned RoomToday, 3:30 PM
+2⭐
📚
Finished HomeworkToday, 5:15 PM
+3⭐
🥦
Ate VegetablesToday, 6:45 PM
+1⭐
🎹
Piano PracticeYesterday
+2⭐
🐕
Walked the DogYesterday
+2⭐

Parenting Made Rewarding

An iOS app that helps parents encourage positive behaviors through a visual star-based reward system. Built end-to-end to keep product instincts sharp.

  • Discovery: Identified gap in parenting apps for positive reinforcement tracking through competitive analysis
  • UX: Designed user flows, information hierarchy, and intuitive journeys optimized for quick parent interactions
  • Analytics: Behavioral pattern insights to guide parent coaching
  • Status: Currently in alpha, planning App Store launch in 2026
About

Background

Product leader based in Israel with 6+ years building and shipping B2B SaaS from 0-to-1 startup through acquisition to scale. Today I lead product execution on a $50M ARR AI/ML platform serving 500K+ subscriptions, delivering technical launches and commercial releases that drive revenue growth.

I bridge technical depth and GTM execution: designing ML products, structuring commercial models, enabling sales teams, and launching products to market. My platform strategy and partner ecosystem approach became a primary growth vector over time.

Domains

  • AI/ML Products
  • IoT & Cloud Platforms
  • B2B SaaS Platforms
  • Fleet Telematics

Leadership

  • Product Growth & GTM
  • Commercial Strategy
  • Partner Ecosystems
  • Cross-Functional Teams

Technical

  • System Architecture
  • ML Pipelines & API Design
  • Edge/Cloud, AWS
  • SQL, Python

AI & Automation

  • Rapid POC Development
  • Data Analytics & Research
  • Technical Documentation
  • LLM Agents & Workflows

The 0-to-1 Journey

2019

Early Employee #8

0-to-1 startup phase

2019-2020

50K Subscriptions

14-month rapid scale

2020

Acquired by Lytx

$500M revenue leader

2020-Present

$50M ARR Platform

TAM → PM II → Senior PM

Education & Certifications

B.S. Business Administration San Jose State University, 2017
Advanced Product Management Product Experts, 2024
Certified ScrumMaster (CSM) Scrum Alliance, 2018
Contact

Let's Connect

Interested in discussing product strategy, platform architecture, and GTM execution.

Open to product leadership opportunities.

Email LinkedIn Call