cover_letter / art_bx2-tXg1Rns

Dear SoFi Hiring Team, SoFi is doing something genuinely rare in financial services: building a full-stack institution — bank charter, lending, investing, payments infrastructure — and betting that the member relationship, not the product silo, is the durable competitive advantage. That orientation matters to me because it mirrors what I set out to build with Fintellect AI: a platform where retail investors could navigate complex financial decisions through guided, context-aware AI — not a chatbot bolted onto a brokerage account, but an intelligent mentor embedded in the financial journey. When I read the SoFi Coach mandate, I recognized the same problem I have been working to solve. ## Technical and AI Foundation My technical credibility in AI runs deeper than product ownership. In 2004, I hand-coded a backpropagation-through-time neural network in C++ for protein structure prediction — work that became a NeurIPS 2014 accepted paper. In 2025–2026, I rebuilt that system from scratch in PyTorch, scaling from 413 parameters to 8 billion across five architectures (feedforward, GRU, Transformer, ESM-2, multi-task), with MLflow experiment tracking, Optuna hyperparameter optimization, and FastAPI serving — 823 automated tests, six Docker containers. That 19-million-fold parameter scale increase in two decades is a useful frame for how I think about AI capability curves. On the RL side, I built a three-phase post-training workbench covering the full RLHF/DPO pipeline: a Reward Lab for designing and A/B testing reward functions across GSM8K, MATH, HumanEval, and UltraFeedback; a Playground for real TRL-powered GRPO/DPO training with live SSE metric streaming on Apple Silicon and CUDA; and an Arena for head-to-head framework benchmarking across TRL, VeRL, OpenRLHF, and NeMo RL with GPU passthrough in Docker. I implemented twelve RL algorithms — PPO, GRPO, DAPO, REINFORCE, REINFORCE++, RLOO, DPO, SimPO, IPO, KTO, ORPO, SPPO — with standardized throughput, memory, and convergence benchmarking. This is the infrastructure layer underneath the kind of personalized, adaptive AI guidance SoFi Coach will need to improve over time. For Fintellect AI specifically, I architected a RAG retrieval pipeline with ChromaDB vector store, multi-provider LLM orchestration across Claude, GPT-4, and Gemini with fallback routing, structured output validation, and token budget optimization. I built domain-specific conversational agents scoped to distinct financial focal points — each delivering guided, context-aware advisory interactions — and integrated AI-powered charting and macroeconomic analysis tools for asset class selection and risk management. That is the direct technical antecedent to what SoFi Coach needs to become. ## Why This Role My arc — from ML researcher to platform PM at Intuit scale to AI founder in fintech — has been building toward exactly this kind of role. The SoFi Coach mandate sits at the intersection of all three: it requires the technical depth to evaluate and integrate LLM capabilities, the platform PM experience to orchestrate cross-functional execution across a complex multi-product ecosystem, and the product instinct to translate member financial anxiety into features that actually change behavior. What specifically draws me to this role is the transition from reactive reporting to proactive guidance. At Fintellect AI, I learned that retail investors do not lack information — they lack a trusted system that tells them what to do next, in plain language, at the right moment. SoFi has the member data, the bank charter, and the product breadth to build that system at a scale no pure fintech startup can match. The opportunity to define the multi-year strategy for SoFi Coach — and to instrument it with the kind of rigorous evaluation infrastructure I built in aeval — is the kind of 0-to-1 mandate I have been preparing for. ## Selected Relevant Experience - **Fintellect AI — RAG pipeline and multi-provider LLM orchestration:** Architected ChromaDB vector store, Claude/GPT-4/Gemini fallback routing, structured output validation, and token budget optimization; built domain-specific conversational agents delivering guided financial advisory interactions in a mobile-first product. - **Fintellect AI — Go-to-market execution:** Led customer discovery, refined platform based on trader feedback, established influencer partnerships, and executed App Store launch — full 0-to-1 product ownership in a regulated financial domain. - **Intuit — ICE platform scale:** Achieved 275% YoY growth in ICE engagements, scaling to 675M+ in FY23 across QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma; scaled throughput from 6K to 50K TPS via rSocket migration supporting ~1.5M concurrent connections with sub-25ms TP99 — the kind of platform reliability SoFi's real-time financial guidance will require. - **Intuit — Developer platform and cross-functional leadership:** Delivered ICE Self-Service platform (DevPortal, GitOps config, ICE Playground), reducing developer onboarding from 2–3 weeks to minutes; conducted enterprise-wide service language assessment across nine languages presented to CTO — demonstrating the executive communication and cross-functional alignment this role demands. - **aeval — AI model evaluation platform:** Built local-first evaluation platform with five core eval types (factuality, reasoning, instruction-following, safety, code generation), adversarial safety testing with refusal detection, bootstrap confidence intervals, Welch's t-test, and Cohen's d effect size — the statistical rigor needed to measure whether SoFi Coach is actually improving member financial health. - **StreamIO AI — Multi-agent orchestration:** Implemented OpenClaw multi-agent orchestration framework with gateway protocol, subagent delegation, profile management, and session switching across real estate, insurance, health/dental, and financial markets — directly applicable to the cross-product, cross-domain guidance SoFi Coach needs to deliver. - **Splunk — Data-driven product prioritization:** Led query performance optimization achieving up to 10x improvements for Fortune 500 beta customers; designed repeatable RICE-based prioritization framework across three microservice backlogs balancing internal, third-party, and enterprise requirements. ## Closing SoFi's mission — helping members get their money right — is not a tagline; it is a genuine gap in how financial services has historically worked. The industry has sold products. SoFi has the infrastructure, the member trust, and now the AI capability to sell outcomes. I want to be the product leader who defines how that happens through SoFi Coach: building the proactive guidance layer that turns SoFi's data advantage into measurable improvements in member financial health, and doing it with the technical rigor and cross-functional discipline that a platform of 9M+ members demands. I would welcome the opportunity to discuss how my background maps to SoFi's specific roadmap. Respectfully, **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info