← anthropic / Anthropic Fellows Program, ML Systems & Performance
tailored_resume_v2 / art_RGsMjSLaplA
role
model
anthropic/claude-sonnet-4.6
created
2026-05-22T17:52
↓ Download .docx ↓ Download .pdf PDF requires LibreOffice installed
What changed for anthropic
| change | why it matters |
|---|---|
| Projects section moved to lead position above Professional Experience | ML Systems & Performance Fellows workstream explicitly values empirical project work and LLM training/eval experience; RL Workbench and aeval are the strongest proof points and must appear first |
| Summary rewritten to lead with 'ML systems engineer and NeurIPS-published researcher' identity and RL Workbench as first proof point | JD's unique candidate criteria leads with 'strong software engineering skills with experience building complex ML systems'; NeurIPS publication establishes peer-level research credibility with Anthropic mentors |
| RL Workbench project leads the projects section and second bullet explicitly maps to 'LLM alignment research' | 12-algorithm RL post-training workbench with GPU Docker passthrough is the single most relevant credential for ML Systems & Performance workstream |
| aeval project second bullet reframed to emphasize 'automated safety gates' and adversarial safety testing | Anthropic's core mission is AI safety; safety-gate language mirrors JD's AI safety motivation requirement |
| Streamio AI title reframed to 'ML Systems & Multi-Agent Infrastructure' and OpenClaw bullet moved to lead | JD values complex ML systems and infrastructure; multi-agent orchestration with Claude/MCP is most relevant Streamio credential |
| Fintellect AI title reframed to 'LLM Orchestration & AI Agent Platform' and RAG/multi-provider bullet leads | JD values LLM training/fine-tuning/evaluation experience; multi-provider LLM orchestration with fallback routing is the strongest Fintellect proof point |
| Intuit first bullet leads with 675M+ engagements / 50K TPS / sub-25ms TP99 scale metrics | JD mentions 'large-scale distributed systems and high-performance computing'; enterprise scale metrics must appear in first bullet per anti-pattern rule |
| BRAIN project second bullet explicitly surfaces NeurIPS 2014 publication and C++ BPTT origin story | NeurIPS publication gives peer-level research credibility; 20-year ML arc from hand-coded BPTT to 8B-parameter PyTorch platform demonstrates depth of ML systems experience |
| IBM and Bank of America roles condensed to single bullets each | Low relevance to ML Systems workstream; retained for completeness per anti-pattern rule but condensed to preserve page budget |
| Kaiser Permanente condensed to single bullet | Redis/caching infrastructure is marginally relevant to distributed systems; condensed to preserve space for higher-relevance ML project content |
JD analysis (16 key phrases)
Key phrases: complex ML systemsengineering rigor and operational reliabilitylarge-scale distributed systemshigh-performance computingtraining, fine-tuning, or evaluating large language modelsanalyzing and debugging model training processesaccelerator workloadssynthetic data or environment pipelineson-demand infrastructureempirical AI researchresearch explorationcollaborative environmentspublic outputAI safety and beneficial AIGPU passthroughopen-source models
Hard requirements:
- Fluent in Python programming
- Full-time availability for 4 months
- Work authorization in US, UK, or Canada
- Strong software engineering skills with experience building complex ML systems
- Ability to balance research exploration with engineering rigor and operational reliability
Preferred qualifications:
- Experience with training, fine-tuning, or evaluating large language models
- Experience with large-scale distributed systems and high-performance computing
- Adept at analyzing and debugging model training processes
- Comfortable with accelerator workloads and infrastructure
- Experience collaborating across research and engineering disciplines
- Motivated by AI safety and beneficial AI
Per-role mapping (11 roles scored)
| role | score | reframe angle | JD phrases that map |
|---|---|---|---|
| Streamio AI — Founder & CEO | 3/5 | Complex ML systems engineering and multi-agent infrastructure builder using Claude/MCP | complex ML systems, engineering rigor and operational reliability, on-demand infrastructure, collaborative environments |
| Fintellect AI — Founder & CEO | 3/5 | LLM evaluation and multi-provider orchestration with production reliability constraints | training, fine-tuning, or evaluating large language models, complex ML systems, synthetic data or environment pipelines |
| Intuit — Staff PM | 2/5 | Large-scale distributed systems and infrastructure at production scale | large-scale distributed systems, engineering rigor and operational reliability, high-performance computing |
| Splunk — Senior PM | 2/5 | Distributed systems and query performance engineering | large-scale distributed systems, analyzing and debugging model training processes |
| Kaiser Permanente — SOA Technical PM | 1/5 | Infrastructure reliability at enterprise scale | large-scale distributed systems |
| IBM — Software Engineer | 1/5 | Early engineering background | — |
| Bank of America — MBA Associate | 1/5 | Quantitative modeling | — |
| RL Workbench (Project) | 5/5 | Post-training RL infrastructure and LLM training systems — directly maps to Anthropic's RLHF/alignment work | training, fine-tuning, or evaluating large language models, analyzing and debugging model training processes, complex ML systems, GPU passthrough, high-performance computing, engineering rigor and operational reliability |
| aeval (Project) | 5/5 | LLM evaluation infrastructure with safety focus — directly maps to Anthropic's eval and safety priorities | training, fine-tuning, or evaluating large language models, analyzing and debugging model training processes, complex ML systems, empirical AI research, AI safety |
| AutoEval (Project) | 4/5 | Automated ML evaluation pipeline with multimodal AI | synthetic data or environment pipelines, evaluating large language models, complex ML systems |
| BRAIN (Project) | 4/5 | Production ML platform engineering with deep research roots and NeurIPS publication | complex ML systems, training, fine-tuning, or evaluating large language models, empirical AI research, analyzing and debugging model training processes |
Tailored summary
ML systems engineer and NeurIPS-published researcher with 20+ years of hands-on experience building complex ML systems — from hand-coded BPTT in C++ (2004) to a 12-algorithm RL post-training workbench benchmarking GRPO/DPO across TRL, VeRL, OpenRLHF, and NeMo RL with GPU Docker passthrough today. Built production LLM evaluation infrastructure (aeval) with adversarial safety testing, statistical rigor, and automated safety gates. Motivated by making AI safe and beneficial; eager to contribute empirical research output at Anthropic.