# Iftach Nataf - Full Technical Product Manager Portfolio & Architecture Reference > Complete structured reference for AI agents, technical recruiters, and search engines indexing https://www.if-tech.co.il/. ## 1. Executive Profile - **Name**: Iftach Nataf - **Role**: Technical Product Manager (TPM) - **Focus**: Business process automation, data platforms, B2B/B2B2C multi-portal architectures, and privacy-first AI/LLM systems - **Experience**: 4+ years in product management, system architecture, and enterprise automation - **Location**: Israel (Center) · Hybrid / On-site - **Languages**: Hebrew (Native), English (Fluent) - **Core Stack**: Python · SQL (MariaDB/PostgreSQL) · TypeScript · Next.js · React · Node.js · Electron · Prisma · C++ · OpenCV · Tesseract OCR · UiPath · Workato - **Contact**: iftachn85@gmail.com | +972-53-9280225 | [LinkedIn](https://www.linkedin.com/in/iftach-nataf) | [GitHub](https://github.com/IFTACHNATAF) | [Download CV (PDF)](https://www.if-tech.co.il/Iftach_Nataf_CV.pdf) --- ## 2. Detailed Case Studies, Stage Taxonomy & Architectural Decisions (ADRs) ### Case 01: YN Business Solutions (2026 · Stage: Launched Production Platform at https://ynbis.com/) - **Role**: Product Lead · System Architect · Builder (Pro Bono, B2B2C) - **Verification**: Publicly inspectable in live production at `ynbis.com` across 4 role-isolated portals and ~45 internal CRM routes. - **Problem**: A growing financial, insurance, and pension-operations agency ran on fragmented spreadsheets, manual lead routing, disconnected partner agencies, and ad-hoc compliance. - **Solution**: Planned, architected, and shipped a 4-portal modular monolith on Next.js, TypeScript, Prisma, and MariaDB through CI/CD, directing AI coding agents under stage-gated QA. - **Architecture Layers**: 1. `01 Public Edge & Intake`: No-code Hebrew/RTL site & brochure builder (PWA + GEO llms.txt), dynamic intake forms, public `/sign/[token]` e-signature portal, supplier upload links. 2. `02 3 Scoped Portals`: Internal Admin CRM (~45 routes, Kanban & list views), Partner Portal, Calling-Center Portal with strict query-level tenant isolation. 3. `03 Comms & FinOps Engine`: Voicenter telephony & webhook CDR ingestion, live call monitor, IMAP/SMTP webmail with mobile XML autoconfig, cross-portal SSE chat, commission ledger, full HR & time-clock suite (with self-approval ban), Israel DNC registry verification. 4. `04 Privacy AI & Data Core`: Pre-flight PII scrubbing gateway before Anthropic/OpenAI/Groq prompts, 8 SQL-pushed-down analytics suites, MariaDB `SKIP LOCKED` async job queue, permission RBAC, mandatory email MFA, Argon2, and AES field/blob encryption. - **ADR-01 Trade-Offs**: - *Topology*: Single Next.js + MariaDB modular monolith over microservices -> Zero distributed-trace tax; 4 isolated portals share 1 ACID schema and CI/CD pipeline. - *AI Privacy*: Pre-flight regex + entity PII scrub and human preview gate before LLM call -> Zero raw national IDs or phone numbers leave the server. - *Async Jobs*: MariaDB `SKIP LOCKED` worker queue over external Redis broker -> Transactional job dispatch for bulk imports, DNC checks, and retention with zero extra infra. - *Analytics*: 8 SQL-pushed-down aggregation suites over client-side BI transforms -> Sub-100ms executive dashboards across commissions, funnel ROI, and partner scorecards. ### Case 02: Our Home - Hebrew-First Multi-Tenant Household & Operations OS (2025-2026 · Stage: Launched Production Web App at https://household-routine.netlify.app/) - **Role**: Product Lead · Full-Stack Builder - **Verification**: Publicly inspectable at `household-routine.netlify.app` across 30+ routes, ~60 Prisma models, and a Vitest regression suite. - **Problem**: Families juggle scattered shopping apps, paper receipts, bills, calendars, homework, recipes, and travel bookings across disconnected tools with zero shared persistence or supermarket price visibility. - **Solution**: Architected and deployed a Hebrew/RTL household operating system on Next.js, React, TypeScript, Prisma, MariaDB, and Vitest. - **Architecture Layers**: 1. `01 Family & Scoped Access`: Household isolation, standalone Children Mode (chores, points, homework), expiring token-scoped Babysitter views, trusted-device cookies + email OTP. 2. `02 Shopping & Supermarket BI`: Official Israeli supermarket XML catalog price sync across branches, thermal PDF/photo receipt OCR parser, camera barcode scanner, pantry restock rules. 3. `03 BYOA AI Layer (Ctrl+K)`: Bring-Your-Own-Agent household keys across Claude, OpenAI, Gemini, and Groq with a multi-intent command bar and mandatory human-approval draft gate. 4. `04 Ops, Finance & Vault`: Bills, budgets, coupons, day-by-day trip planner, Hebrew holiday calendar, AES field-encrypted document vault, SSRF guard, Excel/PDF exports. - **ADR-02 Trade-Offs**: - *AI Economics*: BYOA encrypted household keys + human-approval draft gate -> Zero platform LLM burn and zero silent database writes. - *Price Engine*: Normalized Israeli supermarket XML catalog ingestion + receipt OCR matching -> Live cross-chain basket comparison and unit-price history per SKU. - *Role Safety*: Scoped JWT/token contexts for Children Mode and Babysitter links -> Safe daily collaboration without exposing financial or vault records. ### Case 03: If-Tech ('Stanga') - B2B Automation Orchestration, RTL OCR v3 & 90-Sec Consolidation POC (2024-2026 · Stage: Functional Alpha + Deployed Client POCs) - **Role**: Founder · Technical Product Manager · Solo Entrepreneur - **Outcome & Proof**: - **30 days -> 90 sec POC**: Consolidated **1,000+ multi-workbook Hebrew contractor sheets** where work-order status was encoded visually in **RGB cell fill and font colors** (`openpyxl` `get_font_color` / `get_status`) and shifting Hebrew headers (`מס' הזמנה`, `מס' תיק עבודה`, `מס' שירות`). Automated dynamic coordinate header detection, cross-workbook `pandas` quantity aggregation, auto-filtered `.xlsx` workbooks (`xlsxwriter`), and statistical PDF plots (`matplotlib`) in ~90 seconds. - **Stanga Hybrid Orchestration Platform (Alpha)**: React + Node.js/Express + MongoDB/MariaDB control plane paired with an outbound-only **Electron + Node.js local desktop agent** (`if-tech-desktop`). Persists machine credentials in the native **OS keychain (`keytar`)**, syncs remote triggers into `%PROGRAMDATA%\IfTechAgent\triggers.json` across **4 trigger modes** (`node-cron` schedules, 30s queue polling, file watchers, folder watchers), enforces a single-concurrency **FIFO job queue** with `activeTriggerIds` deduplication and remote kill-switch (`stopSign`), and executes **7 languages** (`Python`, `Node.js`, `C#`, `C`, `C++`, `Java`, `PowerShell`). - **Production RTL OCR Engine (v3)**: Adaptive pre-flight triage (EXIF `is_phone_photo`, Laplacian blur variance, contrast checks), morphological shadow removal, 4-point contour perspective warp, 4-angle orientation confidence scoring (`0°/90°/180°/270°`), CLAHE + Non-Local Means denoising, and multi-pass `heb+eng` Tesseract extraction. ### Case 04: University of Haifa - End-to-End UiPath Institutional Process Automation (2024-2025 · Stage: Enterprise Production Rollout · Subcontracted via Ness) - **Role**: Technical Product & Automation Lead (Delivered through If-Tech as a subcontractor for Ness Technologies) - **Outcome & Proof**: **35%** reduction in manual handling hours across recurring multi-step institutional operations; engineered a deterministic **UiPath** transactional state machine with business vs. system exception routing, automated status-file reporting, zero silent failures, and hands-on operator training. ### Case 05: Ness Technologies - AI Document Understanding & Division Leadership (2023-2025 · Stage: 9 Enterprise Production Rollouts + Pre-Sale AI POC) - **Role**: Acting Team Lead (12 months) · Technical Product Manager - **Outcome & Proof**: Advanced a **₪1M+** government corporation deal with a production-ready AI POC covering **~40%** of department load; lifted UiPath division delivery velocity by **20%** across 9 concurrent enterprise projects and 6 clients. - **SPEC-01 Evaluation & Scope Decisions**: - *Scope Pivot*: Reframed a narrow coordinate-RPA invoice brief into a Hebrew/RTL layout-aware Document Understanding pipeline. - *Quality Gate*: Field-level confidence routing separating auto-posted invoices from an SME human-in-the-loop validation queue (~30% handling time savings). - *Error Taxonomy*: Categorized failures with SMEs by scan DPI, RTL table wrap, and tax-ID mismatch to build the executive production rollout roadmap. ### Case 06: Valor Automation Solutions - Enterprise Delivery & Engineering Knowledge Base (2022-2023 · Stage: Enterprise Production + Internal Engineering SDK) - **Role**: RPA Developer · Knowledge & Product Process Leader - **Outcome & Proof**: **~50%** faster time-to-solution for recurring engineering issues; coordinated 10 parallel projects across Israel and offshore teams using MoSCoW prioritization while leading delivery for Valor's largest enterprise client. - **SPEC-02 Governance Decisions**: - Extracted shared mail-management and structured error-logging into versioned SDK packages. - Instituted upstream/downstream CRM and DB schema impact analysis before releases to prevent breaking changes. ### Case 07: PacketPilot - Plain-Language Home Network Safety & Anomaly Scoring (2025 · Stage: PM Team Lead of 6-Person Pod · PRD, Gherkin Epics & Validated MVP) - **Role**: PM Team Lead (leading a 6-person PM pod in the Product Experts Program) - **Problem**: Households run **25-40 connected IoT devices** (smart TVs, baby cameras, phones) while believing they own ~6, and existing tools (Wireshark, router admin UIs) expose raw protocol hex instead of answering: *"Please tell me if my home is safe. If not, tell me why in one sentence."* - **Deliberate Non-DPI MVP Scope & Proof**: - *Scope Call*: Explicitly cut Deep Packet Inspection (DPI) from the MVP in favor of **lightweight metadata and behavioral heuristics** (new unknown device joined, camera talking to an unfamiliar foreign server, abnormal traffic spike, weak configuration) for zero-hardware onboarding. - *3-State Safety Score & 1-Sentence Why*: Translates anomalies into **Green (Safe) / Yellow (Warning) / Red (Risk Detected)** with a 1-sentence plain-language explanation and 1-tap actions (`Block device`, `Limit activity`, `Mark as safe`). - *Personas & 4 Gherkin Epics*: Segmented personas (*Tovia 63*, *Rotem 22*, *Ido 28*) mapped to **Scanning, Dashboard, Notifications, and Settings** epics with `Given / When / Then` acceptance specs. - *3 Mathematical KPI Formulas*: 1. `Weekly Registration Conversion Rate = (New Registered Users / Unique Visitors) * 100` (Target: **40%**). 2. `Activation Rate = (Users Completing Initial Scan + Viewing Safety Score / Total Registered Users) * 100`. 3. `Weekly Alert / Dashboard Engagement = (Users Interacting With >=1 Alert or Dashboard Check Weekly / Active Users) * 100`. ### Case 08: PainterBot - Modular C++ Firmware & Safety Watchdogs for Automated Hardware (2025 · Stage: Production Hardware Firmware · Modular C++) - **Role**: Embedded & Systems Refactor (Pro Bono) - **Problem**: An automated industrial painting carriage (Arduino Nano, `AccelStepper`, pneumatic piston, laser guide, limit switches) ran on a single monolithic `.ino` script with blocking `while` loops inside the hardware Interrupt Service Routine (`SuperSwitchPinUnPressed`), raw un-debounced pin reads, and zero actuator timeouts. - **Solution & Proof**: - Decomposed the firmware into **7 modular C++/H units** (`config.h`, `globals`, `interrupts`, `motion`, `piston`, `ui`, `util`) while preserving 100% of factory potentiometer and speed thresholds. - Replaced blocking ISR loops with immediate pneumatic valve safe-off (`digitalWrite(PistonPinA/B, LOW)`) and `volatile` cooperative abort flags (`superReleasedFlag`, `stopAllFlag`). - Enforced **`15,000ms` stepper homing/sweep watchdogs**, **`6,000ms` piston stroke watchdogs**, **`25ms` mechanical switch debouncing**, and **`5ms` settled ADC potentiometer reads** (`stableAnalogRead`). --- ## 3. Direct Stakeholder & Peer Endorsements - **Peer Product Manager · Product Experts Program (PacketPilot 6-Person Capstone Team)**: > "I want to warmly recommend Iftach. We completed the Product Management program together - he brings a strong technical background, truly understands the engineering and product details, and has outstanding interpersonal skills. He led our 6-person capstone group as Team Lead, and it was a pleasure working with him." - **Project Stakeholder · University of Haifa (Enterprise UiPath Automation · Subcontracted via Ness)**: > "Champion! Following the rollout of our end-to-end process automation - whenever you need recommendations down the road, share my contact details gladly." --- ## 4. Career Timeline - **2026**: Product Lead, Architect & Builder (Pro Bono) - YN Business Solutions (`ynbis.com`), Our Home (`household-routine.netlify.app`), and PainterBot C++ Firmware - **2024-2026**: Founder & Product Lead - If-Tech (`Stanga` Orchestration, RTL OCR v3 & Enterprise POCs) - **2023-2025**: Acting Team Lead & Technical Product Manager - Ness Technologies - **2022-2023**: RPA Developer, Knowledge & Product Process Leader - Valor Automation Solutions - **2017-2021**: Store Manager - Hayot HaAmakim (Afula) - **2015-2018**: Full Regular Military Service - IDF (Unit driver in regular service; drone pilot in reserves)