← adobe / Principal Product Manager
tailored_resume_v2 / art_CBL6NUouV7c
↓ Download .docx ↓ Download .pdf PDF requires LibreOffice installed
What changed for adobe
| change | why it matters |
|---|---|
| Summary rewritten to lead with 'agentic infrastructure' and MCP/agent loop credibility | JD's first sentence is about owning the agent runtime and harness layer; summary must signal hands-on agentic stack experience immediately |
| Streamio AI title reframed to 'Agent Platform & Developer Tooling' | JD owns 'developer surface for Agent Developers and Agentic App Developers' — title must signal platform ownership, not just 'Founder' |
| Streamio bullets reordered: OpenClaw/MCP leads, control plane second, production app third | JD's top priorities are agent loop contracts, tool protocols (MCP), and control plane ownership — strongest proof points must lead |
| OpenClaw bullet reframed using 'harness layer contracts' and 'agent workflows' language | JD uses 'harness layer' and 'agent loop' as core vocabulary; accurate reframe of multi-agent orchestration work |
| MCP SDK bullet expanded to mention 'skill manifests' and 'tool protocols' | JD explicitly requires fluency in MCP, skills, and tool protocols; candidate's MCP SDK work directly maps |
| Fintellect bullets reframed around 'multi-provider LLM harness' and 'build-vs-adopt' | JD asks PM to 'make architectural calls on what we build vs. adopts from the open ecosystem' — multi-model routing is the proof point |
| Intuit lead bullet kept as scale proof (675M+, 50K TPS) but second bullet reframed around 'developer surface' and 'local dev/prod parity' | JD requires enterprise platform scale credibility AND developer surface ownership; both must appear in first two bullets |
| RL Workbench project moved to lead the projects section and reframed as 'agent quality benchmarking' and 'override and feedback loop' | JD states 'the override and feedback loop is the moat; you'll design it' — RL Workbench's Reward Lab is the exact proof point |
| aeval project reframed around 'enterprise trust & safety guarantees' and 'automated safety gates' | JD requires 'enterprise trust & safety guarantees' as a core platform contract; aeval's adversarial safety testing and safety gates map directly |
| AutoEval project condensed to single bullet and moved to fourth position | Relevant but less central than RL Workbench and aeval for this role; space optimization |
| IBM and Kaiser condensed to single bullets each | Low relevance to agentic platform role; space needed for high-relevance content; minimum 1 bullet rule maintained |
| Bank of America role removed from experience section | Internship from 2011 adds no signal for Principal PM agentic platform role; space optimization for higher-relevance content |
| Deep Learning Education Platform project removed | Educational content project adds minimal signal vs. RL Workbench, aeval, AutoEval, and BRAIN for this role; space optimization |
| Lawrence Berkeley National Laboratory entry removed from projects | NeurIPS paper already captures the research credibility; LBNL entry is redundant and consumes space |
JD analysis (18 key phrases)
Key phrases: agent looptool protocols (MCP, A2A)skills and knowledge groundingenterprise trust & safety guaranteesoverride and feedback loopdeveloper surfaceAgent DevelopersAgentic App Developerslocal dev/prod paritycontrol plane ownershipauth, entitlements, governanceagent quality benchmarksharness layerdata & control planesplatform contractsagentic stackbuild vs. adopts from the open ecosystemtranslate between research, engineering, and GTM
Hard requirements:
- AI platforms, developer tools, or infra products depth
- Shipped platform contracts (not application PM)
- Fluency in MCP, skills, harness patterns, agent loops, tool use, context engineering, evaluation
- Built or extended something in the agentic stack
- Used Claude Code, Cursor, or equivalent agentic dev environments
- Think in contracts and interfaces — ABI stability, plugin/tool terminology precision
- Can hold positions with engineering leads backed by technical reasoning
- Terse writing, audience-calibrated (exec memo, eng spec, customer doc)
- CS or engineering background
Preferred qualifications:
- Hands-on coding ability
- Experience with MCP server development
- Skill manifest authoring
- Agent quality benchmarking tied to customer outcomes
- Control plane ownership: auth, entitlements, governance
Per-role mapping (10 roles scored)
| role | score | reframe angle | JD phrases that map |
|---|---|---|---|
| Streamio AI — Founder & CEO | 5/5 | Agentic platform builder who shipped MCP-native multi-agent orchestration, developer tooling surface, and control plane (auth/session/credential management) from 0-to-1 | agent loop, tool protocols (MCP, A2A), harness layer, developer surface, control plane ownership, auth, entitlements, governance, agentic stack |
| Fintellect AI — Founder & CEO | 4/5 | Multi-model agent harness with knowledge grounding and domain-scoped skills | skills and knowledge grounding, agent loop, harness layer, build vs. adopts from the open ecosystem |
| Intuit — Staff Product Manager | 5/5 | Enterprise developer platform PM who owned platform contracts, SDK surfaces, and build-vs-adopt decisions at Fortune 50 scale | developer surface, platform contracts, local dev/prod parity, build vs. adopts from the open ecosystem, control plane ownership, translate between research, engineering, and GTM |
| Splunk — Senior Product Manager | 3/5 | Platform PM owning developer-facing query language contracts and microservice orchestration | platform contracts, developer surface, control plane ownership |
| Kaiser Permanente — SOA Technical PM | 2/5 | Enterprise platform infrastructure at regulated scale | control plane ownership, enterprise trust & safety guarantees |
| IBM — Software Engineer | 1/5 | Engineering foundation | — |
| Bank of America — Tech MBA Associate | 1/5 | Quantitative analytical foundation | — |
| RL Workbench — Project | 5/5 | Agent evaluation and feedback loop platform — the exact moat Adobe is building | agent quality benchmarks, override and feedback loop, evaluate, override, and improve in production |
| aeval — Project | 5/5 | Agent evaluation platform with safety gates and statistical rigor | agent quality benchmarks, enterprise trust & safety guarantees, override and feedback loop |
| BRAIN — Project | 3/5 | Research credibility and engineering depth | translate between research, engineering, and GTM |
Tailored summary
Principal PM with 12+ years shipping AI platforms, developer tooling, and agentic infrastructure — from MCP-native multi-agent orchestration (OpenClaw) and RL post-training evaluation workbenches to enterprise developer SDKs scaled to 675M+ engagements at Intuit. Hands-on in the modern agentic stack: built and debugged MCP servers, authored agent skills, designed override and feedback loops, and benchmarked GRPO/DPO across TRL, VeRL, OpenRLHF, and NeMo RL. NeurIPS published. Think in platform contracts and interfaces; have held positions with engineering leads and changed them when the argument was better.