Director, Technical Program Management · LinkedIn · Ex Citibank · Capgemini · Oracle
Sunil Ayyappan.
I've spent seven years at LinkedIn leading the platform
transformations that quietly matter. Foundational migrations,
the ML lifecycle, ads and jobs marketplaces. Today I'm
leading Microsoft Digital Work & Platform. Building the
agent platform that will define how AI does work.
Director · since Jun 2024SF Bay Area7+ yrs LinkedIn
7+
Years at LinkedIn · four levels in one org
~10%
Ads YoY revenue growth · two consecutive years
20+
Product systems migrated to Bing Geo
3-yr
Productive ML platform transformation
I'm a technical program leader who has spent his LinkedIn career taking ambiguous, high-stakes technical spaces and turning them into executable programs that scale across the company.
I started with foundational platform migration. Leading LinkedIn's move from its in-house geo system to Bing Geo after the Microsoft acquisition. Then the Productive Machine Learning platform: a three-year transformation that standardized how LinkedIn teams build, train, deploy, and monitor ML in production.
From there, I led TPM teams across Ads, Jobs, and Feed. Programs tied directly to revenue, engagement, and marketplace health. As Director since June 2024, I now lead Microsoft Digital Work & Platform. Building the agent platform that lets the org create, evaluate, and govern digital workers responsibly.
The common thread: taking ambiguity at the technical frontier and shipping platforms that scale.
Thesis · 02
Platforms beat features. Mechanisms beat heroics.
The TPMs who build both. Quietly. Are the ones who make a company stronger every year.
S. Ayyappan · operating principle
Four levels. One company.
2018 ────────────────────────────────────────────── May 2026I am here · Director, 2 yrs in
Sr. TPM
2018 to 2019
Standardization · AI
Bing Geo Migration
Staff → Sr. Staff TPM
2019 to 2022
Data & AI Infra
Productive ML Platform
Sr. Manager TPM
2022 to Jun 2024
Marketplace & Consumer
Ads Marketplace
Jobs & Feed
Director TPM
Jun 2024 to Present
Microsoft Digital Work · LinkedIn Core AI
Microsoft Digital Work & Platform
Core AI Research
Click any card · 2018 → Present · Six platforms
The programs that shaped the role.
Now leading
P · 01 · 2025 to Present
Microsoft Digital Work & Platform
Director TPM · Microsoft Digital Work · Project Aura
Microsoft's agent platform for digital workers. A zero-to-one org anchored by Project Aura, spanning Microsoft and LinkedIn. Includes agent primitives, shared infra, evals & sim, governance, end-user surfaces, and the Forward Deployed Engineer (FDE) team embedded with startups and enterprise customers.
Agent platformFDEProject AuraMicrosoft
P · 02 · Jun 2024 to Present
Core AI Research
Director TPM · regulatory readiness & frontier AI
Director of TPM for Core AI Research. Areas: DSA / DMA / EU AI Act regulatory readiness, human annotation quality, generative recommenders, evaluation rigor.
Responsible AIEU AI ActGenerative recs
P · 03 · 2023 to 2024
Jobs Marketplace & Feed
Sr. Manager TPM · consumer engagement portfolio
TPM execution across two of LinkedIn's most important consumer surfaces. Programs improving sessions, applies, WAU, engagement, recommendation quality, and marketplace health.
Consumer growthRankingMarketplace
P · 04 · 2022 to 2024
Ads Marketplace
Sr. Manager TPM · revenue program team
Led the TPM team across auction, targeting, bidding, budget pacing, experimentation, and marketplace health. Contributed to ~10% YoY revenue growth for two consecutive years.
MarketplaceAuction systems~10% YoY × 2
P · 05 · 2019 to 2022
Productive Machine Learning Platform
Staff → Sr. Staff TPM · 3-year transformation
Three-year company-wide effort to standardize LinkedIn's end-to-end ML lifecycle. Migrated 20+ product teams from fragmented workflows onto a shared production-grade platform.
ML platformMLOps20+ teams
P · 06 · 2018 to 2019
Bing Geo Migration
Sr. TPM · 20+ product systems · post-Microsoft
LinkedIn's enterprise-wide migration from an in-house geo platform to Microsoft Bing Geo. A year-long cross-company program touching profile, jobs, ads, search, recruiter, sales, feed ranking, localization, compliance, analytics.
Platform migrationPost-acquisitionCross-company
§ 05 · InteractivePowered by Claude
Ask my portfolio anything.
A conversational way to dig into my career, programs, or how I think about TPM work. The answers come straight from the source material on this page.
Try a question, or pick one above. Conversations stay on this page.
↗ Live answers · Claude haiku-4.5Rate-limited per visitor
Roles & scope.
Director, Technical Program Management
@ Microsoft Digital Work & Platform · within LinkedIn · Mountain View, CA
Jun 2024 to Present
Lead the TPM team building Microsoft Digital Work & Platform. The agent platform for digital workers, anchored by Project Aura and spanning Microsoft and LinkedIn. Includes a Forward Deployed Engineer (FDE) team embedded with startups and enterprise customers. Previously led Core AI Research within this Director role.
01Drive strategic programs across regulatory readiness (DMA, DSA, EU AI Act), human annotation quality, agent creation, evaluation, simulation, deployment, observability, governance, and generative recommender systems.
02Partner with research, product, engineering, legal, policy, trust, and security to translate AI and digital regulation into engineering programs with owners, milestones, and operating mechanisms.
03Lead the agent-platform TPM team building the full lifecycle for digital workers. From agent creation through evals, simulation, deployment, monitoring, and governance.
04Took over a post-reorg TPM team; ran transparent performance calibration, promoted 3 TPMs Senior → Staff, hired 3 Sr. Staff TPMs.
Sr. Manager, Technical Program Management
@ LinkedIn · Ads Marketplace, Jobs Marketplace, Feed · Mountain View, CA
2022 to Jun 2024
Led TPM teams across LinkedIn's revenue and consumer-engagement surfaces. Ads Marketplace, Jobs Marketplace, and Feed.
01Drove initiatives across auction, targeting, bidding, budget pacing, and experimentation. Contributing to ~10% YoY ads revenue growth for two consecutive years.
02Led Jobs & Feed programs improving sessions, revenue, WAU, engagement, and recommendation quality.
03Built execution mechanisms across product, engineering, data science, ranking, infrastructure, and business teams.
Led a three-year company-wide ML platform transformation that standardized LinkedIn's end-to-end machine learning lifecycle.
01Built and scaled platform capabilities across data pipelines, feature generation, training, experimentation, inference, deployment, monitoring, and health assurance.
02Migrated 20+ product teams across recommendations, search, jobs, ads, feed, and member understanding onto the shared platform.
03Established operational standards for ML production readiness. Release gates, rollback paths, drift detection, operational ownership.
Sr. Technical Program Manager
@ LinkedIn · Standardization AI · Mountain View, CA
2018 to 2019
Joined LinkedIn with 16 years of industry experience; led the post-acquisition Bing Geo Migration.
01Led LinkedIn's enterprise-wide migration from an in-house geo system to Microsoft Bing Geo across 20+ product and platform systems.
02Coordinated product, engineering, legal, data, and Microsoft partner teams to migrate taxonomy, APIs, services, pipelines, and reporting flows.
03Preserved product relevance, compliance, member experience, and business continuity through a year-long, high-risk platform change.
Staff Technical Program Manager
@ Citibank · New York, NY
2016 to 2018
Operated at the intersection of technology, finance, and operations to modernize enterprise AI and data systems.
01Led AI investment planning and execution across personalization, fraud detection, and compliance platforms.
02Built financial and operational models that improved infrastructure efficiency by 30%, unlocking $10M+ in annual savings.
03Partnered with engineering and risk leadership to align spend, capacity, and regulatory outcomes.
Sr. Technical Program Manager
@ Capgemini · Global banking clients
2012 to 2016
Supported global banking clients through analytics-driven transformation.
01Built operating models and execution plans for large, multi-stakeholder initiatives across financial-services clients.
02Delivered planning frameworks and performance tracking for data and analytics programs at scale.
Tech Lead
@ Oracle · Mission-critical financial platforms
Earlier · 2002 to 2012
Tech-led globally distributed teams delivering mission-critical financial platforms. Where I developed the operating discipline that has shaped every program I've led since.
01Managed globally distributed teams shipping mission-critical financial platforms.
02Developed early operating discipline across large engineering programs in regulated environments.
Six numbers I'm proud of.
~10%
Ads YoY revenue growth × 2 years
Ads Marketplace TPM team
20+
Product systems migrated
Bing Geo · post-acquisition
3-yr
ML platform transformation
Productive Machine Learning
20+
Teams adopted shared ML lifecycle
Recs · Search · Jobs · Ads · Feed
Jun '24
Promoted to Director
Now leading Microsoft Digital Work
6
Domains led at LinkedIn
Geo · ML · Ads · Jobs · AI · Agents
Leadership
TPM org design & hiring
Coaching ICs to Staff & beyond
Executive engagement (VP+)
Cross-company alignment
Performance turnaround
Execution
Platform migration at scale
Strategy → roadmap operationalization
Adoption-led program management
Risk & dependency management
Experimentation discipline
Domains
Foundational & applied AI / ML platforms
Agent platforms & evals
Marketplace & auction systems
Consumer growth & ranking
Responsible AI & regulation
Post-acquisition integration
Stanford GSB
LEAD. Corporate Innovation
MIT
Data Science & Big Data Analytics
University of Madras
Master's in Computer Science
01
Launching AI Products at Scale at LinkedIn
Predictive Analytics World (PAW) for Business
2021
02
Launching AI & Data Programs
Grace Hopper Celebration · Bangalore (co-author)
2021
03
Women in AI
LinkedIn · co-led · 220+ external attendees
2020
04
Springboard Rise · Career panel
LinkedIn · audience of 400
2020
05
Jira Portfolio & the Ratchet tool. Scaling TPM craft
LinkedIn TPM Forum / Foundation Lunch & Learn
2020
"Among all TPMs I closely worked with during the past 7+ years since I joined LinkedIn, Sunil is the best so far. For many times, I felt he worked like my right arm."
"I have worked with many TPMs at Google and LinkedIn and would rank Sunil very high in my list. His technical abilities and willingness to dive deep are above and beyond."
"Sunil is the sort of TPM you can trust with big projects. From proposal to driving progress to providing updates at the right times, his execution is quite mature."
A year-long post-acquisition platform migration touching 20+ product systems. The hardest kind: foundational, invisible when it works, catastrophic when it doesn't.
Role
Sr. TPM
Period
2018 to 2019
Scale
20+ systems
§ 01
Context
After Microsoft acquired LinkedIn in 2016, LinkedIn needed to rationalize several foundational platform capabilities with Microsoft's broader ecosystem while preserving its product experience, member trust, and business continuity. LinkedIn's in-house geo system was deeply embedded across the company's product and platform surfaces. Geo data powered profiles, jobs, ads targeting, search relevance, recruiter workflows, sales, feed ranking, localization, compliance, and analytics.
This was not an API replacement. It was a cross-company platform migration touching 20+ product systems, multiple data contracts, downstream product assumptions, legal and privacy considerations, and member-facing experiences. The core challenge was migrating from one canonical understanding of geography to another without breaking product semantics, search relevance, targeting accuracy, reporting continuity, or member trust.
§ 02
Scope
End-to-end execution across product, engineering, legal, data, infrastructure, and Microsoft partner teams. The work included:
Mapping LinkedIn's internal geo taxonomy to Bing's entities. Countries, regions, cities, postal areas, aliases, localized names, ambiguous locations.
Identifying every product dependency on geo data across LinkedIn's ecosystem.
Creating migration plans for 20+ product and platform teams.
Aligning product managers on user-facing impact, edge cases, acceptance criteria, and rollout sequencing.
Partnering with legal and privacy to ensure geographic data handling met regional obligations.
Coordinating engineering across data pipelines, APIs, services, search indexes, ad targeting, analytics, and UI.
Defining validation, parity checks, fallback mechanisms, and release gates.
Managing executive communication, risk tracking, escalation, and milestone governance over a year-long program.
§ 03
Technical complexity
Geo was a foundational primitive, not an isolated feature. A single location entity could influence: a member's profile location, job search radius and recommendations, recruiter candidate discovery, sales region segmentation, ads targeting and auction eligibility, feed personalization, search ranking, compliance experiences, and analytics reporting.
The migration required strong controls around taxonomy parity, data quality, localization, backward compatibility, and phased rollout. It also required careful handling of historical data, because reporting and experimentation systems often depend on consistent identifiers over time.
The success criteria were never just shipping code. They were preserving trust across hundreds of downstream use cases.
§ 04
Why this matters
This program demonstrated my ability to lead high-risk, cross-company platform migrations where the success criteria were not just shipping code, but preserving trust across hundreds of downstream use cases. It required technical depth, stakeholder alignment, legal review, dependency management, and change management. Operating across boundaries that don't share the same incentives.
Platform migrationPost-acquisitionCross-companyLegal & complianceDependency mgmtLeading without authority
P · 05 · 2019 to 2022 · Staff → Sr. Staff TPM
Productive ML Platform.
A three-year transformation. Moving LinkedIn from team-specific ML workflows to a reusable, scalable, production-grade ML lifecycle platform.
Role
Staff → Sr. Staff TPM
Period
3-year transformation
Adoption
20+ teams
§ 01
Context
At LinkedIn, machine learning was deeply embedded across recommendations, search, feed ranking, jobs, ads, notifications, member understanding, growth, and trust. Before this platform effort, different teams had their own workflows for data preparation, feature generation, training, experimentation, deployment, and monitoring. Creating fragmentation, duplicated infrastructure, inconsistent reliability, and slower velocity.
Training infrastructure. Common workflows for offline, recurring, and distributed training; versioning and reliability.
Feature management. Consistency between offline training and online serving.
Experimentation. Integrated with A/B platforms so teams could evaluate model impact before broad rollout.
Inference and serving. Predictable latency, scalability, reliability in production.
Deployment. Standardized promotion from dev → staging → prod, with release gates, rollback paths, and readiness checks.
Health assurance. Model health checks, drift detection, production alerts, operational ownership.
Adoption. Worked with 20+ product teams to migrate ML development and deployment onto the platform.
§ 03
Why this was hard
This was a three-year transformation, not a launch. Phase one built the platform foundation. Phase two drove adoption across LinkedIn's product ecosystem.
The hardest part wasn't technical. It was changing how teams built ML. Product teams already had local workflows, custom assumptions, and production pressures. My role required a platform strategy that balanced central standardization with product-team flexibility, and defined adoption sequencing that respected each team's ML maturity.
Production ML is not just "train a model and deploy it." It requires data engineering, repeatability, governance, deployment automation, monitoring, and operational ownership.
§ 04
What I drove
Understood how each team built ML for recommendations, search, jobs, ads, feed.
Identified common lifecycle patterns; converted team-specific work into reusable platform capabilities.
Defined adoption milestones and migration sequencing.
Built onboarding paths for teams at different ML-maturity levels.
Created executive visibility into progress, risks, business impact, adoption, reliability, developer productivity.
Established operational standards for ML production readiness.
ML platformMLOps3-year transformationAdoption-led20+ teamsPlatform thinking
P · 04 · 2022 to 2024 · Sr. Manager TPM
Ads Marketplace.
A real-time system balancing advertiser value, member experience, marketplace efficiency, and long-term ecosystem health.
Role
Sr. Manager TPM
Outcome
~10% YoY × 2
Surfaces
Auction · bidding · pacing
§ 01
Scope
Over two years, my team partnered with product, engineering, data science, and business teams across the LinkedIn Ads Marketplace. Contributing to approximately 10% year-over-year revenue growth for two consecutive years.
Auction systems. Improved which ads are eligible, ranked, priced, and shown.
Targeting. Better advertiser reach for the right professional audiences while maintaining member relevance and policy.
Bidding. Improved bid strategy, advertiser optimization, delivery, and marketplace efficiency.
Budget pacing & allocation. How advertiser budgets are distributed across time, campaigns, placements.
Marketplace health. Balanced growth with relevance, quality, latency, advertiser ROI, member trust, fairness.
§ 02
Why this was hard
Ads marketplace work is hard because changes often affect multiple competing metrics. A change that improves short-term revenue can hurt advertiser ROI. A change that improves delivery can harm member experience. A change that improves relevance can reduce liquidity. A change that increases auction pressure can affect pricing and retention.
My role was to help the org manage this complexity through clear prioritization, experimentation discipline, cross-functional alignment, and principled execution.
Manage the tradeoffs. Build the mechanisms. Let the experiments decide.
§ 03
What I led
Built program plans across auction, bidding, targeting, budget delivery, experimentation.
Aligned revenue, product, engineering, and data science stakeholders.
Defined success metrics across revenue, advertiser value, member experience, marketplace health.
Tracked initiative- and portfolio-level impact.
Managed tradeoffs between growth, quality, reliability, and advertiser trust.
Created visibility for leadership on risks, dependencies, expected outcomes.
MarketplaceAuction systems~10% YoY × 2ExperimentationCross-functionalRevenue program
P · 03 · 2023 to 2024 · Sr. Manager TPM
Jobs & Feed.
Two of LinkedIn's most important member-engagement and monetization surfaces. ML, social-graph signals, content quality, freshness, personalization, trust, engagement optimization. All at once.
Role
Sr. Manager TPM
Surfaces
Jobs · Feed
Theme
Member engagement
§ 01
Jobs Marketplace
Programs aimed at increasing jobs sessions, job-seeker engagement, apply starts and completions, recommendation relevance, recruiter and employer outcomes, marketplace liquidity, jobs-product revenue, search and recommendation quality, and experiment velocity. The work required balancing job-seeker needs, employer outcomes, marketplace supply and demand, monetization, and member trust.
§ 02
Feed
Initiatives that improved WAU, feed sessions, content engagement, ranking quality, recommendation relevance, creator and professional community activity, notification and re-engagement loops, trust and quality signals, and long-term retention. Feed is complex because it combines ML, social-graph signals, content quality, freshness, personalization, trust, and engagement optimization in one surface.
§ 03
What I led
Drove program execution across Jobs Marketplace and Feed initiatives.
Partnered with product, engineering, data science, design, and business teams.
Tracked metric movement across engagement, revenue, relevance, marketplace health.
Improved planning discipline across multiple product areas.
Helped teams move from idea → experiment → rollout.
Built mechanisms for dependency management and leadership visibility.
Many initiatives had dependencies across ranking models, data pipelines, experimentation platforms, product surfaces, and monetization systems. The work was to create structure across that portfolio.
Operating at the intersection of frontier AI capability development, responsible AI, regulatory readiness, product impact, and platform execution.
Role
Director TPM
Since
Jun 2024
Focus
Frontier · Regulation
§ 01
Regulatory context
The regulatory environment for large digital platforms and AI systems changed significantly. The EU Digital Services Act introduced rules for online services used by European citizens. The EU AI Act entered into force on August 1, 2024, with most obligations becoming fully applicable on August 2, 2026. The Digital Markets Act focuses on contestability and fairness for large gatekeeper platforms.
Against this backdrop, AI program leadership requires not only technical delivery but governance, transparency, documentation, risk management, evaluation, and responsible deployment.
§ 02
DMA, DSA & EU AI Act readiness
I lead TPM work across regulatory readiness programs. Translating requirements into executable technical programs with legal, policy, engineering, AI, data, trust, and product teams.
Requirement interpretation with legal and policy.
Technical impact assessment.
Product and platform dependency mapping.
Data access, explainability, transparency, governance workflows.
Risk management and executive reporting.
Mechanisms for sustainable compliance. Not one-time response.
§ 03
Human annotation quality
Annotations influence model performance, safety, relevance, and fairness. I lead programs to improve annotation quality, scale expert and crowd workflows, build better guidelines and calibration loops, reduce labeling ambiguity, and create feedback loops between model errors and annotation strategy.
§ 04
Generative recommenders
Programs where traditional recommendation approaches are enhanced by generative AI. Focus areas: relevance, personalization quality, latency and serving constraints, evaluation methodology, responsible AI review, experimentation, production rollout readiness.
Translate broad strategic themes. "AI regulation," "annotation quality," "generative recommenders". Into concrete execution plans with owners, milestones, risks, metrics, and operating mechanisms.
Responsible AIEU AI ActDSA / DMAAnnotationGenerative recsFrontier AI
P · 01 · 2025 to Present · Director TPM · Now leading
Microsoft Digital Work & Platform.
A zero-to-one org at Microsoft, anchored by Project Aura. The enterprise-scale agentic workforce platform. Spans Microsoft and LinkedIn and operates like a startup inside the company.
Role
Director TPM
Anchor
Project Aura
Org
Microsoft Digital Work
§ 01
The teams I lead
The org is structured around the parts of the agent lifecycle. Each team has its own bar, its own rhythm, but they ship as one platform.
Forward Deployed Engineers (FDE). Embedded with startups and enterprise customers; the team that turns the platform into shipped customer value and feeds learnings back into product.
Agent primitives. The reusable building blocks for agent creation, tools, memory, and orchestration.
New agent onboarding. Repeatable processes for launching agents without rebuilding the system.
§ 03
Why this is harder than traditional platforms
Agent behavior is probabilistic, tool-using, context-dependent, and often multi-step. Success requires stronger mechanisms for evaluation, simulation, observability, safety, and human oversight than traditional software platforms ever needed.
The TPM team coordinates across AI engineering, platform engineering, product, security, legal, enterprise systems, data and evaluation teams, FDE, internal customer teams, and executive stakeholders on both sides of the joint effort.
Agent platforms aren't software in the old sense. They're operating systems for delegated work. And that means the TPM bar is higher, not lower.
§ 04
What I lead
Establish the founding operating model: team shape, planning rhythms, execution standards for a zero-to-one platform spanning Microsoft and LinkedIn tech stacks and compliance surfaces.
Ship the first wave of platform bets. New products in close partnership with startups and enterprise customers, from agent primitives and shared infra to end-user surfaces. While keeping delivery predictable under high ambiguity.
Single connective layer across agents and teams; own end-to-end execution health across LLM integration, agent orchestration, inference infra, evaluation pipelines, and Responsible AI governance.
Coach the FDE team on customer-facing technical judgment, escalation, and feedback loops back into the platform roadmap.
Agent platformProject AuraFDEMicrosoft Digital WorkZero-to-oneGovernanceEvals & sim
