Program & Project Manager
Complex initiatives,delivered end to end.
Program and project manager with 6+ years delivering technology programs at Applied Materials, IBM and Keshet Media Group.

©2026
/Project rescue · Fractional PM · AI governance · IT & infrastructure programs
(01) About
Hey!
I work at the intersection of software, AI adoption, product delivery, and cross-functional operations.

With a hands-on technical foundation and experience at Applied Materials, IBM, and Keshet Media Group, I take fragmented processes and align R&D, IT, Security, and Finance to a single cadence.
Results from recent roles
40+
AI workflows moved under IT/R&D security compliance
Keshet Media Group
7
Redundant tools cut through unified AI tool licensing
Keshet Media Group
900
Project boards archived or consolidated in a platform cleanup
Keshet Media Group
−30%
Compute cost after moving servers to a colocation site
Applied Materials
1,200+
Employee facility whose IT infrastructure was modernized
Applied Materials
0
User disruption across system migrations
IBM
Services
How I can help
For technology projects that are late, over budget or unclear, and for technology people need to understand. Five ways to work together.
(02)
Project rescue
A technology project that’s late, over budget or stuck, and nobody can say exactly why.
What you get
- A health check in the first week
- A reset plan the sponsor signs off
- Weekly status that shows real risk
Done before: Caught a budget gap of millions of dollars before it became an overrun (Applied Materials).
Usually 1–3 months
Fractional PM / PMO
Teams that need a senior project or program manager, but not a full-time hire.
What you get
- Roadmap, plan and risk register
- One cadence across R&D, IT, Security and Finance
- A live dashboard for leadership
Done before: Ran an IT and R&D portfolio of 10+ projects a year for 4,000+ employees (Applied Materials).
A few days a week, month to month
AI governance setup
Companies where people adopt AI tools faster than IT and Security can approve them.
What you get
- A map of the AI tools and workflows in use
- A simple approved route and policy
- A licence and cost review
Done before: Moved 40+ unapproved AI workflows into approved, secure tools (Keshet Media Group).
Usually 4–6 weeks
IT & infrastructure programs
An office build-out, compute expansion or system migration with many vendors and long lead times.
What you get
- One schedule and critical path
- Long-lead procurement tracking
- Change control with a budget reserve
Done before: Delivered a 16-month AI compute build a month early and about 5% under budget (Applied Materials).
The length of the program
Explainer & training videos
Teams whose new hires, customers or non-technical staff need to understand complex technology quickly.
What you get
- Short animated videos in 2D or 3D
- Hebrew or English
- A script your expert signs off before production
Done before: Built a Hebrew 3D onboarding explainer on how chip fabs find defects (concept demo).
About 2 weeks per video
Case studies
Selected work
(03)

Keshet Media Group
2026 / Case 01
Shadow AI Mitigation & Executive Steering Framework
Key impact: Moved 40+ unapproved employee AI workflows into secure, compliant tools. The CIO and I chose the tools; I owned the route to them and the governance around it.
- AI Governance
- Security Boundaries
- SaaS License Optimization
Explore full case study

Applied Materials
2026 / Case 02
Multi-Year Infrastructure Modernization (1,200+ Employees)
Key impact: Managed the multi-year IT/R&D infrastructure portfolio for a 1,200+ employee facility, building automated SQL & Tableau dashboards for live C-suite visibility.
- CapEx/OpEx
- SQL & Tableau Telemetry
Explore full case study

IBM
2023 / Case 03
Multi-System Migration & Zero-Downtime Hypercare
Key impact: Led cross-system migrations and disaster-recovery initiatives with hands-on UAT ownership, achieving zero operational downtime for end users.
- UAT & Hypercare
- Disaster Recovery
Explore full case study
More work
Applied Materials · 2023–2026
An AI compute farm, a month early and 5% under budget
Problem
- R&D’s AI chip simulation and testing was outgrowing the compute it had.
What I built
- A water-cooled compute farm of 150 racks on a 400G network, delivered over 16 months as the lead IT PM.
How it works
- Gear with 12–18-month lead times ordered at first design; change control and a 5–10% reserve absorbed a mid-project jump to 40 kW racks.
What changed
- Live a month early, about 5% under budget, with 99.99% uptime after go-live.
Applied Materials · 2023–2026
A budget gap caught before it was an overrun
Problem
- The annual operating plan underestimated the real cost by millions of dollars.
What I built
- A corrected cost, budget and scope baseline, worked out with Finance.
How it works
- Raised the gap early with the numbers, sized its impact with Finance, and updated cost, budget and scope together.
What changed
- The revised budget was approved, so the gap never became a mid-year overrun.
Keshet Media Group · 2026
An approved route for vibe-coded apps
Problem
- Employees were shipping AI-built apps fast; leadership couldn’t see what existed, who owned it or what data it touched.
What I built
- A standard route from request to production for apps employees build with AI.
How it works
- A short requirements doc, architecture review, cloud registration, a CI/CD pipeline, and automated code and security scans.
What changed
- Leadership sees what’s built, who owns it and what data it uses, and building moved off personal tools into controlled ones.
Keshet Media Group · 2026
From scattered card spend to one corporate card
Problem
- Departments bought on their own credit cards; charges arrived with no invoice or cost center, leaving gaps at month-end close.
What I built
- A procure-to-pay process with one corporate card, standardized with Finance.
How it works
- Every purchase is linked to an invoice and to the buying department’s cost center.
What changed
- One corporate card and a clear budget split per department.
Keshet Media Group · 2026
One source of truth for people and AI agents
Problem
- Knowledge was scattered across docs, boards and chats and left with people; employees’ AI agents ran with no company context.
What I built
- Started a company knowledge base that serves both employees and the AI agents they build.
How it works
- Systems, owners, dependencies and processes mapped into standard wiki templates through each department’s COO; each agent gets only its creator’s access.
What changed
- Agents built on it start from current company facts instead of data re-entered by hand.
Applied Materials · 2023–2026
IT for a new 12-floor tower
Problem
- A new tower with offices, R&D labs and a clean room needed all of its IT from scratch.
What I built
- End-to-end IT delivery for the building, from empty floors to handover.
How it works
- A project charter, design reviews at 30/60/90%, critical-path planning and contractor oversight, bridging Facilities, engineering and IT.
What changed
- All 12 floors handed over, on new network and physical-security standards I defined.
Keshet Media Group · 2026
A helpdesk dashboard that exposed the gaps
Problem
- No view of service requests: tickets with no assignee, WhatsApp requests never categorized, and no SLA.
What I built
- A helpdesk dashboard showing every request, its status and who has it.
How it works
- Brings tickets into one view and flags those with no category or owner.
What changed
- Adopted by the team with positive feedback, and it set the next fix: categories, owners and SLAs.
Buildots · 2020
Monitoring a production AI pipeline
Problem
- An AI pipeline processed site data for many customer projects at once, and machine-learning errors had to be caught fast.
What I built
- NOC monitoring of the production pipeline, as NOC Engineer.
How it works
- Tracked each project’s runs, resolved errors, escalated to R&D, and set the outsourced QA team’s daily priorities.
What changed
- Resolution metrics and fixes fed back to R&D, improving the pipeline in production.
Projects I’ve Built
Real organizational pain. Real solutions.
(04)

Coded project · Keel
2026 / Live demo
One app instead of MS Project, Monday and Slack
Program management software for multi-team tech programs. The schedule, the boards and the team conversation live in one place, and AI turns emails and chat into plan changes the PM approves, so nobody chases status updates by hand.
Explore the project
3 connected projects · Discover, decide, sustain
Citizen AI Governance
People will build and use AI on their own. Governance has to make the approved route the easy one, from idea to production.
Stage 1 · Discover
Status: PlannedShadow AI Discovery & Amnesty
Finds the AI tools people already use on personal accounts and unregistered licenses, triages each one (stop, migrate or formalize), and offers a no-blame way to self-report.
Stage 2 · Decide
Status: LiveInnovation & PRD Marketplace
An internal portal where any employee turns an idea, a problem or a vibe-coded prototype into a policy-checked PRD with an AI wizard, then publishes it to a marketplace where managers prioritize it and assign it to a builder.
Stage 3 · Sustain
Status: LiveCitizen App Registry & Handover
A registry for employee-built apps: stages from idea to production, a named maintenance owner on every app, and live repo health signals from the GitHub API.
I sit between the people who build technology and the people who fund and depend on it. I turn requirements into plans, constraints into decisions, and status into foresight leadership can plan with.
How I think
A few things I believe about product
(06)
01
Early product work should expose the riskiest assumption, not prove the easiest one.
Technical feasibility is almost never the assumption most likely to kill adoption. Whether the user who matters most actually has the problem you think they have usually is. I design the first weeks of any new product around stress-testing that assumption, not building the most demoable version.
Read moreShow less
Technical feasibility is almost never the assumption most likely to kill adoption. Whether the user who matters most actually has the problem you think they have usually is. I design the first weeks of any new product around stress-testing that assumption, not building the most demoable version.
02
Value isn’t complete until users can perceive it.
In complex categories like cybersecurity, AI and B2B platforms, a product can work perfectly and still feel like nothing is happening. The product job isn’t done until value is legible: visible enough to trust, specific enough to recommend, real enough to renew.
Read moreShow less
In complex categories like cybersecurity, AI and B2B platforms, a product can work perfectly and still feel like nothing is happening. The product job isn’t done until value is legible: visible enough to trust, specific enough to recommend, real enough to renew.
03
Design is upstream, not downstream.
The most important design decisions aren’t about color or polish. They’re structural: what appears first, what’s hidden, what’s the default, what the system implies about how users should behave. These choices shape behavior before anyone reads a single word, and they’re product strategy, not styling.
Read moreShow less
The most important design decisions aren’t about color or polish. They’re structural: what appears first, what’s hidden, what’s the default, what the system implies about how users should behave. These choices shape behavior before anyone reads a single word, and they’re product strategy, not styling.
04
I’d rather build something a smaller group loves than something everyone tolerates.
In early products, love is a more reliable signal than completeness. It tells you whether you’ve found something worth scaling, or just something that technically works. Minimum lovable, not minimum viable.
Read moreShow less
In early products, love is a more reliable signal than completeness. It tells you whether you’ve found something worth scaling, or just something that technically works. Minimum lovable, not minimum viable.
Writing
Notes from the field
(07)
Sep 2026 • AI product
The AI Integration Trap: Why More Automation Doesn't Mean Better Products
Most AI features don't fail because of the model. They fail because they take the wrong amount of control. A 1978 framework for choosing how much autonomy AI should have.
Read
Sep 2026 • AI practice
How to Avoid Slop (Content, Visual, Ideation), and Why You Need to Ask the LLM to Argue With You
Why default AI output trends toward generic, agreeable slop, and the technique (asking the model to argue against you) that fixes it.
Read
7 articles
All posts