← cerebrassystems / AI Models, Product Manager
tailored_resume_v2 / art_qIIPIBr2eTo
role
model
anthropic/claude-sonnet-4.6
created
2026-05-22T15:39
↓ Download .docx ↓ Download .pdf PDF requires LibreOffice installed
What changed for cerebrassystems
| change | why it matters |
|---|---|
| Projects section promoted to appear before Professional Experience | aeval and RL Workbench are the strongest proof points for Cerebras' core requirement of model quality evaluation frameworks and systematic benchmarking — leading with them maximizes perceived fit before the reader reaches work history |
| Summary rewritten to lead with 'model strategy, evaluation frameworks, and developer ecosystems' framing | JD's first ownership area is 'Strategic Model Portfolio' and second is 'Product Quality & Customer Success' via evaluation frameworks — mirroring this language immediately signals fit |
| aeval project promoted to lead the projects section (previously listed second) | aeval's 5 eval types, adversarial safety testing, and statistical rigor directly satisfy the preferred qualification 'experience writing model quality evaluations and system prompt harnesses' — strongest single proof point for this role |
| RL Workbench bullet reframed to connect benchmarking discipline to 'proving production-grade inference performance on Cerebras hardware' | JD asks PM to 'design benchmarks and evaluations that prove our models deliver production-grade performance' — explicit bridge makes the connection legible to a non-technical recruiter |
| Streamio lead bullet reframed around OpenClaw agentic flows and chat completions API | JD hard requirement: 'expertise on agentic flows' and 'comfortable using Python with the chat completions API' — OpenClaw is the strongest proof; leading with it satisfies both |
| Fintellect lead bullet reframed as 'multi-provider LLM orchestration' and 'model portfolio management' | JD's core ownership is the model portfolio across frontier and open-source models — multi-provider Claude/GPT-4/Gemini orchestration with fallback routing is the closest analog from candidate's experience |
| Intuit lead bullet kept as 675M+ / 50K TPS scale metric | JD implies success metrics around inference scale and platform adoption — burying enterprise scale is an anti-pattern; metrics belong in first 2 bullets |
| Splunk lead bullet reframed around 10x performance optimization mapping to inference speed tradeoffs | JD requires 'balance tradeoffs between quality, latency, throughput, and cost' and 'select and prioritize performance optimizations' — 10x query perf improvement is the closest analog |
| BRAIN project reframed to foreground PyTorch, 8B parameter scale, and NeurIPS credential | JD preferred qual: 'knowledge and passion for open-source models and generative AI research' — NeurIPS publication and PyTorch hands-on work satisfy both the research credibility and technical ecosystem requirements |
| Deep Learning Education Platform project removed to save space | Weakest fit for this role; space better used for aeval/RL Workbench detail which are directly relevant to model evaluation requirements |
| IBM and BofA roles condensed to single bullets each | Low relevance to AI Models PM role; retained to satisfy 5+ years SWE hard requirement and avoid gaps, but condensed to preserve page budget for high-relevance content |
JD analysis (20 key phrases)
Key phrases: model portfoliofrontier and open-source modelsday0 launchesmodel quality evaluationssystematic evaluation frameworksinference speedquantizationspeculative decodingagentic computationchat completions APIPyTorchHugging FacevLLMSGLanggo-to-marketproduction-grade performancemodel optimizationtechnical contentdeveloper ecosystemcross-functional leadership
Hard requirements:
- 5+ years PM experience at Senior PM level or above
- 5+ years total technical work experience (SWE, ML researcher, solution engineer)
- Knowledge of open-source models and generative AI research
- Knowledge of PyTorch, Hugging Face, vLLM, SGLang ecosystem
- Comfortable using Python with chat completions API for model testing
- Ability to thrive in fast-paced entrepreneurial environment
- Cross-functional leadership across engineering, marketing, sales
Preferred qualifications:
- PM experience at model training lab or company implementing open-source models
- Experience writing model quality evaluations and system prompt harnesses
- Experience with agentic flows and LLM model family architectures
- Experience writing technical marketing assets and social media
- Experience with model optimization/compression (quantization, speculative decoding)
- Understanding of model compilers and optimization
- Contributor to vLLM, SGLang, PyTorch, or Hugging Face communities
- Experience writing application code for code generation or deep research
Per-role mapping (10 roles scored)
| role | score | reframe angle | JD phrases that map |
|---|---|---|---|
| Streamio AI — Founder & CEO | 4/5 | Agentic AI platform builder with hands-on LLM orchestration, multi-agent architecture, and 0-to-1 product execution | agentic computation, chat completions API, go-to-market, model quality evaluations, technical content |
| Fintellect AI — Founder & CEO | 3/5 | Multi-model LLM orchestration and inference optimization for production AI applications | frontier and open-source models, model portfolio, go-to-market, production-grade performance |
| Intuit — Staff Product Manager | 4/5 | Enterprise-scale platform PM with developer ecosystem ownership and cross-functional launch leadership | cross-functional leadership, go-to-market, developer ecosystem, technical decision-making, production-grade performance |
| Splunk — Senior Product Manager | 3/5 | Performance-focused platform PM with rapid delivery track record and structured prioritization | inference speed, production-grade performance, cross-functional leadership |
| Kaiser Permanente — SOA Technical PM | 2/5 | Large-scale infrastructure PM | production-grade performance |
| IBM — Software Engineer | 2/5 | Technical foundation satisfying 5+ years SWE requirement | — |
| Bank of America — Tech MBA Associate | 1/5 | Quantitative background | — |
| RL Workbench — Post-Training RL Platform | 5/5 | Hands-on model evaluation platform builder with multi-framework RL benchmarking — strongest proof of model quality evaluation expertise | systematic evaluation frameworks, model quality evaluations, production-grade performance, PyTorch, quantization, model optimization |
| aeval — AI Model Evaluation Platform | 5/5 | Production model evaluation platform with statistical rigor — strongest proof of evaluation framework design | model quality evaluations, systematic evaluation frameworks, production-grade performance, benchmarks and evaluations |
| BRAIN — Protein Structure Prediction ML Platform | 3/5 | Published ML researcher with hands-on PyTorch model development from small to 8B parameter scale | PyTorch, frontier and open-source models, model optimization |
Tailored summary
AI product leader with 12+ years driving model strategy, evaluation frameworks, and developer ecosystems at scale — from hand-coding BPTT in C++ (2004) to building RL post-training workbenches that benchmark GRPO/DPO across TRL, VeRL, OpenRLHF, and NeMo RL today. Built production model evaluation platforms (aeval) with systematic quality frameworks across factuality, reasoning, safety, and code generation. Deep expertise in agentic flows, multi-model LLM orchestration, and inference optimization; scaled platform infrastructure to 675M+ engagements at Intuit. NeurIPS published researcher; UC Berkeley Engineering + CMU MBA.