cover_letter / art_tPdtyru804I

Dear Databricks AI Hiring Team, Databricks sits at a rare intersection: the company that gave the world Apache Spark, Delta Lake, and MLflow is now defining how enterprises build, deploy, and operate AI at scale. That combination of infrastructure depth and AI ambition is exactly the kind of platform challenge I have spent the last several years working toward — from shipping developer frameworks that scaled to 675M+ engagements at Intuit, to building my own RL post-training workbench that benchmarks GRPO, DPO, and PPO across TRL, VeRL, OpenRLHF, and NeMo RL today. ## Technical Foundation My AI/ML work is not adjacent to my product work — it is the same work. In 2004 I hand-coded a neural network in C++ with custom backpropagation through time to predict protein secondary structure; that system was accepted at NeurIPS 2014. In 2026 I rewrote it as a production PyTorch platform spanning five architectures (feedforward, GRU, Transformer, ESM-2, multi-task) and 413 to 8B parameters — a 19-million-fold scale increase — with MLflow experiment tracking, Optuna hyperparameter optimization, and FastAPI serving across six Docker containers. More directly relevant to Databricks AI: I built a full RL post-training workbench covering the complete RLHF/DPO pipeline. The Reward Lab lets users design and A/B test reward functions (RLVR, learned, hybrid) across GSM8K, MATH, HumanEval, and UltraFeedback. The Playground runs real TRL-powered GRPO and DPO training with live SSE metric streaming on Apple Silicon (MPS) or CUDA. The Arena benchmarks TRL, VeRL, OpenRLHF, and NeMo RL head-to-head with GPU passthrough in Docker containers. I implemented 12 RL algorithms — PPO, GRPO, DAPO, REINFORCE, REINFORCE++, RLOO, DPO, SimPO, IPO, KTO, ORPO, SPPO — with standardized throughput, memory, and convergence benchmarking across frameworks. This is the kind of first-principles technical depth that lets a PM go deep with research engineers rather than just relay requirements. I also built aeval, a local-first model evaluation platform with five core eval types (factuality, reasoning, instruction-following, safety, code generation), adversarial safety testing with refusal detection, and statistical rigor: bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, and saturation detection. The stack — FastAPI orchestrator, TimescaleDB, Redis job queue, Next.js dashboard, Ollama — mirrors the kind of evaluation infrastructure that enterprise AI platforms need to make model quality measurable and repeatable. ## Why This Role The arc from NeurIPS researcher to developer platform PM to AI founder maps cleanly onto what Databricks AI is building: foundational capabilities for agents, model development, and complex workflow orchestration. I have lived on both sides of that boundary — as the engineer who implements the algorithms and as the PM who has to make them accessible to millions of users. What specifically draws me to this role is the mandate to define capabilities that do not exist yet. Databricks is not filling in a feature matrix; the team is determining what enterprise AI development looks like for the next decade. The combination of agent orchestration, model training, and data platform integration is exactly the problem space I have been working in — building OpenClaw, a multi-agent orchestration framework with gateway protocol and subagent delegation, and architecting RAG retrieval pipelines with multi-provider LLM orchestration, fallback routing, and structured output validation at Fintellect AI. ## Selected Experience - **RL Post-Training Workbench (2026):** Implemented 12 RL algorithms with cross-tab workflow lineage tracking and head-to-head framework benchmarking (TRL, VeRL, OpenRLHF, NeMo RL) with GPU passthrough in Docker — directly applicable to Databricks' model development and fine-tuning surface. - **aeval Evaluation Platform (2025–2026):** Built CI/CD-integrated model evaluation with regression detection and automated safety gates — the kind of trust infrastructure enterprise customers need before deploying AI at scale. - **Intuit ICE Platform — 275% YoY growth, 675M+ engagements (FY23):** Scaled platform throughput from 6K to 50K TPS via rSocket migration supporting ~1.5M concurrent connections with sub-25ms TP99 across QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma. - **Intuit Developer Onboarding (Java/Python SDK Starter Kits):** Reduced developer onboarding from 2–3 weeks to minutes in pre-prod and under 24 hours for production, while mitigating $1M+ in projected opex growth — a direct analog to Databricks' developer experience mission. - **Splunk Search Orchestration (Go microservices, SPL/SPL2):** Delivered Scheduler Service end-to-end in approximately four months; led query performance optimization achieving up to 10x improvements for a Fortune 500 beta customer — demonstrating the ability to ship in fast-moving, competitive environments. - **OpenClaw Multi-Agent Orchestration:** Designed and implemented gateway protocol, subagent delegation, profile management, and session switching — enabling coordinated AI agent workflows across multiple industry verticals. - **SQL and Usage Data Fluency at Intuit:** Used SQL and BigQuery to prioritize developer pain points across approximately 20 mobile apps and 30+ product SKUs; built Asterias, a declarative asset lifecycle management platform with GraphQL API. - **NeurIPS 2014 — Protein Structure Prediction:** Published research on artificial neural networks for secondary structure prediction, grounding my AI product instincts in research-grade technical understanding. ## Closing Databricks' mission — enabling every organization to harness the power of data and AI — is not a tagline I am drawn to abstractly. It is the specific problem I have been working on from multiple angles: as a researcher, as a platform PM at scale, and as a founder building AI products from zero. I want to bring that full stack of experience to a team that is defining what enterprise AI development looks like for the next generation of builders. I would welcome the opportunity to discuss how my background maps to the Databricks AI roadmap. Sincerely, **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info