The Person
Who is Farid Boutaghane
Know the person before the meeting starts.
FB
Farid Boutaghane
Chief Technology Officer · Philip Morris International · Lausanne, Switzerland
Infrastructure, cloud architecture, and integration sit under Farid's remit — including cybersecurity. He was the executive sponsor behind PMI's all-in AWS migration: 400 applications moved in 2 years (2020–2022), with all data centers sold before the migration started. No going back — literally.
He spoke at AWS Cloud Day Switzerland 2023 — his highest-profile public appearance — on how technology drives PMI's smoke-free transformation with "tangible outcomes beyond cost savings." That framing (ROI first, tech second) is his lens. He does not speak at dozens of conferences. He is deliberate and outcome-oriented.
He spoke at AWS Cloud Day Switzerland 2023 — his highest-profile public appearance — on how technology drives PMI's smoke-free transformation with "tangible outcomes beyond cost savings." That framing (ROI first, tech second) is his lens. He does not speak at dozens of conferences. He is deliberate and outcome-oriented.
"Due to the shared accountability model of the technology stack, our teams can focus more on essential workloads rather than compliance."
— Farid Boutaghane, AWS Migration Case Study
"Cost savings will become more tangible as we continue modernizing. AWS helps us achieve carbon reduction goals without straining resources."
— Farid Boutaghane, AWS Migration Case Study
How to open the conversation
Reference AWS Cloud Day Switzerland — he invested time speaking there, so it's a credible signal you did your homework. Lean into "tangible outcomes" language. He doesn't want to talk about technology for its own sake. He wants to know what changes in the business. Start there.
Market Intelligence
What We Know About PMI's Data Platform
Research-based intelligence from public sources — November 2024 disclosures and leadership interviews.
190+
Data engineers on the platform across 25 markets
38.6M
Connected IQOS devices generating live behavioral telemetry
95
Markets governed by PMI's GenAI Hyperscaler Platform (launched 2024)
2027
Target for global Distributed Data Network rollout across all markets
Distributed Data Network (DDN) — PMI's Data Mesh in Motion
PMI published a detailed technical disclosure in November 2024. They've moved from centralized data warehouses to a federated data mesh: 30+ global nodes, each an independent AWS account, sharing data via Snowflake without duplication. By 2027, every market will be live on this architecture. Data engineering stack: Snowflake · Databricks · Matillion · Atlan · Profisee · AWS S3 · Kafka.
GenAI Hyperscaler Platform — Centralized AI Governance at Scale
Mid-2024, PMI stood up a centralized hub to govern all AI development across 95 markets. Five use case tracks: content creation, knowledge management, AI assistants, ideation & insights, digital personas. Microsoft Copilot rolled out to ~20,000 employees. 3,500+ staff enrolled in AI literacy program. "Responsible AI" is a formal workstream — they won't rush and they won't cut corners.
Knowledge Graph + Information Fabric — The 2025–2027 Roadmap
PMI's next architectural layer is an Information Fabric: metadata-driven process lineage linked to domain metadata. On top: a Knowledge Graph (Altair Graph Studio + AWS Neptune) to enable GenAI grounded in enterprise data. This is where Databricks Unity Catalog, AI Functions, and Mosaic AI become directly relevant.
Key Talent Signal — Chief Data Strategist Has Left
Parviz Shariff, who designed PMI's Distributed Data Network and co-authored the AWS DDN blog post, has since left PMI for Coca-Cola HBC. The person who built the DDN strategy is gone. His successor is not yet publicly named. This may create an open door on the architecture and strategy side.
Tech Stack
Cloud
AWS (primary)
Azure AD (identity)
Multi-cloud
Data Platform
Snowflake
Databricks ✓
Matillion
Atlan
Profisee (MDM)
BI / Analytics
Power BI
Amazon Redshift
Jupyter
Storage / Streaming
Amazon S3
Apache Kafka
Amazon Kinesis
Amazon RDS
MongoDB
Graph
Altair Graph Studio
AWS Neptune
AI / Prod.
M365 Copilot
GenAI Hyperscaler (internal)
Aera Decision Cloud (supply chain)
"Data and every technology that's related to data is our focus area. The combination of supercomputing with machine learning is like the invention of the steam engine or electricity."
— Michael Voegele, Chief Digital & Information Officer, PMI
Strategic Fit
Where Customertimes Can Help
Four angles — each grounded in PMI's actual situation, not a generic pitch.
1
Visionary & Architecture Advisory
Consulting · Strategic Support
PMI Signal
- Chief data strategist (Parviz Shariff) has departed — DDN architecture leadership gap
- Information Fabric + Knowledge Graph roadmap is defined but the implementation path is open
- Farid's language: "digital maturity at all levels" — he thinks about platform and people, not just tools
- No public Databricks reference story — an "untold story" gap that CT could help fill
CT Fit
- CT designs medallion architectures on Databricks across CPG, HLS, and manufacturing at scale
- Can advise on Databricks Unity Catalog + Delta Sharing as the governance layer across DDN nodes
- Can co-design the GenAI layer on top of the Knowledge Graph — grounding LLMs in governed lakehouse data
- Potential to co-develop a PMI + Databricks reference architecture story for joint publishing
2
Project Execution — Data Activation & ML Operationalization
Project Work · Implementation
PMI Signal
- 38.6M connected devices generating continuous behavioral telemetry — needs ML pipelines at very high throughput
- GenAI use cases across 95 markets: content, knowledge, personalization — need Databricks-backed feature stores
- DDN onboarding all markets by 2027 — each node needs standardized data engineering patterns
- Supply chain AI (Aera, 25,000+ materials) — decision-layer integrations needed
CT Fit
- CT builds Bronze/Silver/Gold medallion pipelines with CDC at scale — 190+ data engineers would benefit from standardized quick-starts
- MLflow + Databricks Model Serving for production ML from IQOS telemetry
- CT's Databricks + Snowflake integration expertise aligns with PMI's polyglot stack
- BI activation (Power BI on top of Databricks) — CT has delivered this pattern at multiple clients
3
Data Activation Across PMI Systems
Use Case Discovery · Data Products
PMI Signal
- PMI shifted from B2B → direct-to-consumer — new data flows across retail, digital, and device channels
- Consumer engagement platform live in 20 markets — analytics + personalization demand
- Legacy application elimination (1,000 of 2,600) creates data migration and modernization demand
- IQOS behavioral data largely untapped for cross-market AI — still mostly per-market silos
CT Fit
- CT's Sales Data Products framework (from TE Connectivity) maps directly to consumer engagement data activation
- Legacy-to-AI modernization: CT Migration Accelerator handles ABAP→Databricks; SAS→Python/Databricks playbook exists
- CT builds cross-domain data products (Finance 360, Customer 360) that span legacy + modern systems
- PMI's Snowflake + Databricks polyglot model is exactly the architecture CT has productized
4
Staff Augmentation — Databricks-Certified Engineers
Staff Aug · Capability Extension
PMI Signal
- 190+ data engineers across 25 markets — a federated team that needs specialized Databricks skills
- 500+ employees earned AWS certs in 2 years — PMI invests in skill-building, open to augmentation
- Data strategy leadership gap (Shariff's departure) suggests need for senior technical voices
- DDN rollout to all markets by 2027 is a surge capacity problem — they will need external capacity
CT Fit
- CT provides Databricks-certified engineers across data engineering, MLflow, and Unity Catalog
- CT staff aug is embedded — not a body shop. Engineers work inside PMI's delivery model
- Flexible engagement: sprint teams, PoC pods, or long-term embedded capacity
- CT has augmented 25+ global enterprise data teams across CPG and healthcare verticals
Use Cases
Data Activation Across PMI Systems
Concrete workloads to bring to the conversation — grounded in PMI's public roadmap.
🔗 IQOS Telemetry → Personalization ML
38.6M connected devices. Puff frequency, session length, device health — all streaming continuously. Today: per-market analytics. Opportunity: a unified Databricks Feature Store + MLflow serving layer that powers personalized consumer journeys across all markets. CT deliverable: end-to-end medallion pipeline (Kafka → Bronze → Silver → Gold → Feature Store → Model Serving).
🌐 DDN Node Onboarding Accelerator
PMI needs to onboard every market by 2027. Today it takes significant configuration per node. Opportunity: a CT-built templated Databricks workspace provisioning framework integrated with Matillion ETL patterns — so new market onboarding goes from weeks to days. Aligns directly with PMI's stated goal of "hours of configuration" per node.
🧠 GenAI Grounded in the Knowledge Graph
PMI's 2025–2027 bet: Knowledge Graph (Altair + Neptune) + Information Fabric → GenAI. The gap: connecting LLMs to governed, lineage-tracked enterprise data. CT solution: Databricks Unity Catalog as the semantic governance layer, Mosaic AI as the inference layer, RAG pipelines grounded in Delta Lake. CT built this for healthcare and manufacturing clients.
🏭 Legacy Application Data Liberation
1,000 apps being eliminated. Many hold dark data — operational history, consumer records, transactional data that should live in the lakehouse. CT's Migration Accelerator automates ABAP→Databricks; CT's SAS→Python/Databricks playbook handles analytics migrations. This is a repeatable factory, not a one-time project.
🛒 Consumer 360 Data Product
PMI is now direct-to-consumer in 20 markets. Consumer data sits across IQOS devices, digital engagement platforms, retail POS, and compliance systems. CT's Customer 360 data product framework on Databricks creates a unified, governed, AI-ready consumer profile — powering the personalization and digital personas use cases from PMI's GenAI strategy.
📦 Supply Chain Intelligence Layer
PMI runs Aera Decision Cloud across 25,000+ materials, 40+ factories. Aera makes decisions — but the underlying ML models need fresh, governed data to be trustworthy. CT builds the data pipeline layer that feeds decision intelligence platforms: clean Silver/Gold layers, model-ready features, lineage tracking. 70% Aera adoption, 40% planner workload reduction is the current benchmark. CT can push it further.
Why Customertimes
What We've Done — Relevant Proof
Not case studies. Signals. Reference clients you can name in the room.
TE Connectivity
OneData platform on AWS + Databricks. Finance 360, Sales Data Products, Procurement Analytics — all on medallion architecture with Delta Lake. 4 cross-domain data products, unified governance across business units. Same federated challenge PMI faces.
Abbott EPD
Asia-Pacific analytics platform on Databricks: Delta Lake, DBT, Unity Catalog. Consumer healthcare — same industry DNA. Data products feeding commercial teams across multiple markets. PMI's market node architecture mirrors this.
IQVIA
Patient services platform and iEngage (field rep engagement) on Databricks. Healthcare data at scale with strict governance requirements — analogous to PMI's regulatory exposure across 95 markets.
CT's Databricks Capabilities — What's Actually Differentiating
- Medallion Architecture Factory — ingestion, CDC, Bronze/Silver/Gold as a standardized, repeatable delivery pattern. This is what PMI needs for DDN node onboarding.
- Unity Catalog Governance — data classification, lineage, access control. PMI's Atlan + Profisee sit above the data layer; Unity Catalog governs inside the lakehouse.
- AI Activation Layer — MLflow, Feature Store, Model Serving. CT ships ML from prototype to production, not just builds the pipeline.
- CT Migration Accelerator — AI-powered ABAP→Databricks migration tool. LLM-based code translation. PMI's 1,000-app elimination creates direct demand.
- Snowflake ↔ Databricks Bridge — CT has productized the integration between Snowflake (PMI's structured data hub) and Databricks (ML workloads). This is PMI's exact polyglot architecture.
- CPG Industry Depth — CT brings domain knowledge in consumer goods, commercial analytics, and supply chain — not a generic SI applying a data template.
The Conversation
20-Minute Flow
Suggested structure. Adapt in real time based on where Farid takes it.
-
0–3 minOpen — acknowledge what they've builtReference the AWS Cloud Day Switzerland talk. "Tangible outcomes beyond cost savings" — that framing stood out. The DDN architecture you published in November is one of the most detailed data mesh case studies in the FMCG space. We've been following it closely." This signals preparation and earns credibility without being sycophantic.
-
3–7 minProbe — where Databricks sits in their current realityAsk: "We know Databricks is in your stack. Help me understand — is it primarily for ML workloads today, or are you also running data engineering pipelines through it?" And: "With Parviz moving on, has the Knowledge Graph + Information Fabric roadmap changed shape, or is that still the 2025–2027 priority?" These questions show you know the details. Let him talk.
-
7–13 minPosition — where CT can specifically contributePick 2 of the 4 angles based on what he surfaced: (1) Architecture advisory and the GenAI/Knowledge Graph layer — CT can be a thinking partner, not just an implementer. (2) The DDN node onboarding factory — if they're onboarding 5+ markets per year through 2027, they need a repeatable pattern. That's a CT deliverable. (3) Staff aug with Databricks-certified engineers who already know the CPG domain. Name TE Connectivity and Abbott as comparators.
-
13–18 minConcrete — propose a small sharp entry pointSuggest a 2-week architecture assessment: review current Databricks usage, identify gaps against their 2027 roadmap, and produce a prioritized map of where CT can accelerate. Low commitment, high signal. Alternatively: a 4-week PoC on one of the 6 use cases — ideally the IQOS telemetry → personalization use case, which has a clear outcome and is not politically sensitive.
-
18–20 minClose — one clear next stepAsk: "Who else on your team should be part of a next conversation — would you include the data engineering lead or is this a CTO-level conversation first?" Then: "Can we send you a 1-page scope for the assessment within the week?" One ask, one commitment. Don't overstuff the close.
Questions to Ask
- "Where does Databricks sit in your delivery priorities vs. Snowflake right now?"
- "The Information Fabric roadmap — is the KG layer still the 2025 focus, or has the priority shifted?"
- "What's the hardest part of onboarding new markets to the DDN?"
- "Is the GenAI Hyperscaler Platform running on top of the lakehouse, or is it more disconnected today?"
- "Where are your biggest skill gaps on the Databricks/ML side — internal team or third-party?"
Things to Avoid
- Don't lead with CT's company overview — he doesn't have time for that
- Don't say "we can do everything" — pick a lane and own it
- Don't pitch staff aug first — it's in the agenda, but open with vision and capability; staff aug is the logical follow-on
- Don't mention tobacco or smoke-free transition unless he brings it — it's a charged topic and irrelevant to the data conversation
- Don't overpromise on timelines — Farid drives by "tangible outcomes" and will hold you to specifics
The One Thing to Leave Him With
Customertimes has built the Databricks delivery muscle across complex, multi-market enterprise environments in consumer goods and healthcare. We're not a generalist SI — we have an industry-specific delivery pattern that maps directly to PMI's DDN architecture, and we can operate at the intersection of consulting (the Knowledge Graph vision) and execution (the node onboarding factory). We want to be your Databricks implementation partner for the 2027 buildout.