application_answers / art_BIiU8p8InnA

Content

{
  "answers": [
    {
      "question": "Do you know anyone currently at Glean?",
      "answer": "No, I do not currently know anyone at Glean."
    },
    {
      "question": "Are you willing and able to commit to the hybrid policy if hired? (MUST BE SF BAY AREA BASED)",
      "answer": "Yes \u2014 I am based in Oakland, CA, which is in the SF Bay Area, and I am fully able to commit to four days per week in the Mountain View office."
    },
    {
      "question": "How did you hear about Glean?",
      "answer": "Job Board"
    },
    {
      "question": "Please input the total years of experience you have that are relevant for this role.",
      "answer": "12+ years. This includes 3+ years as Staff PM at Intuit owning developer platforms and AI-powered infrastructure at enterprise scale (675M+ engagements, 50K TPS), 2 years as Senior PM at Splunk owning search orchestration microservices and analytics products, 6+ years as SOA Technical PM at Kaiser Permanente, and concurrent hands-on AI product work building RAG pipelines, multi-agent orchestration frameworks, RL post-training workbenches, and model evaluation platforms as a founder."
    },
    {
      "question": "What AI tools are you currently using today and how are you using them?",
      "answer": "I use AI tools both as a builder and as a daily practitioner. On the development side, I use Claude (via the MCP SDK) as the core reasoning engine in StreamIO AI \u2014 it performs real-time contextual analysis of screen captures, generates Pine Script for trading, and powers automated real estate CMA reports via Redfin/Zillow APIs. I built an OpenClaw multi-agent orchestration framework on top of it with gateway protocol and subagent delegation. For Fintellect AI, I run multi-provider LLM orchestration across Claude, GPT-4, and Gemini with fallback routing and a ChromaDB RAG retrieval pipeline. I also built aeval, a local-first model evaluation platform using Ollama, FastAPI, TimescaleDB, and Redis, with adversarial safety testing and statistical rigor (bootstrap CIs, Welch's t-test, Cohen's d). For my RL Workbench, I use TRL-powered GRPO and DPO training with live SSE metric streaming and benchmark 12 RL algorithms across TRL, VeRL, OpenRLHF, and NeMo RL. In my day-to-day product craft, I use Claude and GPT-4 for PRD drafting, user story generation, competitive analysis synthesis, and rapid prototyping of interaction flows \u2014 treating them as a thought partner for moving quickly from ambiguous problem spaces to concrete product specs."
    }
  ],
  "skipped": [
    "First Name",
    "Last Name",
    "Email",
    "Phone",
    "Resume/CV",
    "Cover Letter",
    "LinkedIn Profile",
    "Website"
  ]
}