Founding Engineer / Hands-on CTO

Moshe Harush

I get complex systems under control and keep the product moving.

I learn the full system, stabilize production, and modernize the parts that block product delivery. More than 14 years of hands-on work across backend, architecture, infrastructure, product delivery, and technical leadership.

View selected work Email me

Senior Tech Lead at Selectika AI · Building UnderGrid

Moshe Harush outdoors in Tel Aviv
Architecture · Delivery · AI Tel Aviv, Israel

  • Architecture

    Map system boundaries, constraints, data flows, and failure modes.

  • End-to-end leadership

    Lead engineering from product decisions through production operations.

  • Applied AI

    Run, fine-tune, evaluate, and integrate local and hosted models in production systems.

Selected work

What I owned, what I changed, and why.

Three examples of end-to-end ownership: platform modernization, a digital health product, and high-traffic consumer systems.

Public scope

The source and systems remain private or proprietary. These case studies cover only the architecture, decisions, and outcomes I can share. They include no internal code, customer data, traffic figures, or sensitive identifiers. Any diagrams are sanitized and any example data is synthetic.

Selectika AI · Platform modernization · 2024–present

Modernizing a multi-tenant platform without pausing active deploys.

Context
A memory-intensive monolith, blocking workloads, and per-client file configuration no longer matched the product's multi-tenant direction.
What I owned
Hands-on ownership of backend architecture, infrastructure, deploys, production, and the modernization path.
Key decisions
Kubernetes-based workload separation, database-driven tenant configuration, RabbitMQ-backed asynchronous processing, and incremental replacement.
Current result
Long-running work moved out of request paths. Tenant onboarding moved from manual file changes across deployments to a database-driven workflow. Workloads gained clearer scaling boundaries while active deploys continued.

UnderGrid · Digital health · Founder-led product

Building UnderGrid end to end across web, native, backend, infrastructure, and AI.

Context
The product needs to connect coach workflows, member interactions, consented health integrations, and safety-aware AI assistance across web and native apps.
What I owned
Product definition, modular TypeScript architecture, backend, PWA and native delivery, shared contracts, infrastructure, and production operations.
Key decisions
A modular monolith with explicit boundaries, server-side AI tools with validation and confirmation gates, and native health integrations behind consent-aware interfaces.
Outcome
One product system connects web, native, backend, AI tools, and health integrations. Safety and privacy checks are part of the architecture.

Yad2 · High-traffic consumer products · 2018–2020

Moving a critical login flow from PHP to Node.js without a big-bang rewrite.

Context
Critical PHP systems had to be modernized while teams kept shipping high-traffic consumer flows.
What I owned
Engineering leadership for Second Hand, Alerts, and Favorites, plus cross-team work on system behavior and platform dependencies.
Key decision
Migrate incrementally from PHP to Node.js, beginning with a production-critical login flow rather than replacing the whole system at once.
Outcome
The critical login boundary moved to Node.js while product teams kept shipping, creating a repeatable path for the broader migration.

Operating method

Map it. Stabilize it. Then change the parts that matter.

Modernization starts with constraints, rollback paths, and ownership, not with choosing a new stack.

  1. Map the system

    Map boundaries, ownership, dependencies, data flows, and the actual critical path.

  2. Find the actual bottleneck

    Separate product pressure, production failures, team limits, and legacy code that still does useful work.

  3. Stabilize production

    Improve observability and reliability before broadening the change surface.

  4. Create a safe cutover path

    Add rollback points, measurable checkpoints, parallel delivery, and room for gradual replacement.

  5. Move one boundary at a time

    Start where the change removes the most risk or unlocks the most product work.

  6. Prove the new path, then remove the old one

    Verify behavior and metrics, remove the legacy path, and document the new runtime path and ownership.

Applied AI

Models are only one part of shipping AI.

I run, fine-tune, and evaluate local and hosted models, build tool-connected agents, and design the product boundaries around permissions, data, latency, cost, and failure. Evaluation and fallback behavior are part of the product, not an afterthought.

Models, inference & training

Local and hosted inference, fine-tuning and training workflows, embeddings, retrieval, evaluation, and model/runtime trade-offs.

Agents & tools

Tool-connected agents for coding and research, with explicit context, permission boundaries, evaluation, and feedback loops.

Product systems

Schema-validated output, safety checks, human confirmation, logs, cost and latency budgets, and privacy boundaries.

Model changes should not change the product contract, safety checks, or fallback behavior.

Experience

Experience across product, platform, and production.

I stayed hands-on as the scope expanded: code, architecture, infrastructure, hiring, delivery, and production.

Selectika AI

Senior Tech Lead

Primary hands-on engineer across backend, architecture, infrastructure, deployments, and production; leading the platform's multi-tenant modernization while active delivery continues.

Canditech

Full-Stack Developer

Built mobile-first assessment flows, backend capabilities, internal test-execution tools, server-side rendering, and improvements to CI/CD and developer workflows.

Startup Booster

Tech Lead & Strategic Partner

Worked as a hands-on fractional CTO for early-stage products, translating founder requirements into technical plans, production systems, and delivery practices.

Yad2

Team Lead & Web Architect

Led teams across Second Hand, Alerts, and Favorites; drove gradual PHP-to-Node.js modernization, interviewed 30+ developers, and built internal training programs.

WebId Group

CTO, Co-Founder & Full-Stack Developer

Co-founded and scaled a web agency, built a custom CMS that powered 50+ multilingual SEO-focused sites, and automated recurring content and lead-generation work.

IDF · Contact Battalion 372

Logistics management system

Identified an operational gap and independently designed, built, and deployed a logistics management system while serving in a non-technical role.

Technical scope

Architecture & backend
Node.js, PHP, Python, APIs, modular systems, multi-tenancy, event-driven design
Product & leadership
Product engineering, technical strategy, mentoring, hiring, delivery ownership
Platform & operations
Kubernetes, Docker, AWS, CI/CD, observability, production diagnosis
Data & AI
PostgreSQL, MySQL, MongoDB, Redis, Elasticsearch, model runtimes, RAG, agents

Contact

Building a product that needs hands-on technical ownership?

I am interested in founding-engineer and hands-on CTO roles where I can own the path from product decisions to production.

Email me WhatsApp
WhatsApp