Mihir Parmar

San Francisco, CA, United States0

Principal Product Manager | AI Product Builder | Zillow Document Intelligence | Founder, En Route Luxe

I build products and businesses - from AI platforms at enterprise scale to a luxury travel brand built from scratch. My career has followed one thread: find the complexity people work around, and turn it into something they can build on. Trading tech and Regulated platforms at Goldman Sachs, Deutsche Bank, UBS, Credit Suisse. 6.5 years at Zillow shipping document intelligence to production: a multi-tenant platform processing 1M+ documents a year, digital leases that doubled landlord activation, and multimodal LLM document understanding with human-in-the-loop review (accuracy from ~85% to 98%+). All of it built on continuous discovery with the brokers, agents, and transaction coordinators who live in these documents. The turn that made it work was refusing the obvious target. A contract that's 92% right is still unusable if nobody can tell you which 8% is wrong, so we gated on per-field confidence instead of average accuracy. The lesson that shapes everything I build: the model is never the product. The evals, guardrails, and review systems that earn a professional's trust - that's the product. Today that mindset powers En Route Luxe, a boutique luxury travel advisory ($500K+ bookings, grown entirely through repeat bookings and referrals). Luxury should feel personal, not transactional - so behind the human touch runs a virtual team of AI agents I built with Claude Code. The product work runs on it, not just the business: discovery and destination research, itinerary QA, content workflows, with automated guardrails on every itinerary. It is a working skill library, and it is where I learned the part that is harder than building agents - knowing where they add real leverage and where a simpler approach wins. Several workflows are deliberately manual because an agent would not have earned its keep. I call it the editor layer. The through line is the same in both: I build the layer that makes automated judgment trustworthy enough to rely on. At Zillow that was evals and confidence routing on a million documents a year. At En Route Luxe it is guardrails on itineraries where a missed transfer ruins a trip. Same problem, different stakes. What's next: senior product leadership where AI meets high-stakes trust: document intelligence, agentic systems, and travel, an industry I know from the inside. Outside of work: cricket, Bollywood, travel, and the emotional roller coaster of supporting Manchester United. Always happy to connect with people building in AI workflow automation, enterprise AI agents, SaaS, marketplaces, and travel. Visit mihirparmar.ai.

AI Fluency

No AI stories yet

Nothing posted yet.

Experience

Founder

En Route Luxe

Sep 2025 — Present · San Francisco Bay Area

Founded En Route Luxe to make luxury travel feel personal again — crafted around how each client actually wants to experience the world. We work with a select number of clients each year, designing end-to-end trips where every detail is intentional and every contingency accounted for. By the numbers: 20+ clients/year, 50+ trips/year, $500K+ in annual bookings — grown entirely through repeat bookings and referrals. Under the hood, the product work itself runs on an AI stack I built and ship daily: a library of custom Claude Code skills I wrote for recurring workflows (discovery and destination research, itinerary assembly and QA, brand voice, content engine), plus scheduled agent automations. It functions as a working skill and prompt library, applied to a business with paying clients rather than to a demo. The discipline that makes it work: knowing where AI adds real leverage and where it does not. Judgment - supplier selection, disruption response, taste - stays human, and several workflows stay deliberately simple because an agent would not have earned its keep. I write publicly about this as "the editor layer."

Principal Product Manager

Zillow

Sep 2019 — Sep 2026 · San Francisco Bay Area

Built Zillow's document intelligence platform end-to-end — from a reusable multi-tenant foundation, to national-scale proof with Zillow Rentals digital leases, to an AI-first document + workflow product for brokers and agents. It started with discovery: years of working directly with the brokers, agents, and transaction coordinators who live in these documents. • Shipped multimodal LLM document understanding to production. The call that made it work: average accuracy was the wrong gate, because a document that's 92% right is unusable if you can't tell which 8% is wrong. So we routed on per-field confidence with human-in-the-loop guardrails, partnering with engineering on model selection, thresholds, and cost/latency tradeoffs. Accuracy from ~85% (OCR/NLP) to 98%+, with ~100% on reviewed compliance-critical fields, across 1M+ documents a year. • Built the eval practice that gated every rollout: golden datasets seeded from years of review corrections, per-field accuracy gates, and production quality monitoring — the system that turns AI that demos well into AI professionals trust. • Consolidated 20+ point solutions into an AI-first document + workflow product; early pilots showed ~60% less admin time per deal and ~$1M annual savings for a mid-size brokerage. • Zillow Rentals: shipped digital leases nationally in ~12 months; instrumented the lease funnel (create → send → sign) and iterated with experiments — doubled landlord activation and grew market share 18% → 30%. • Scaled the multi-tenant Document Platform across 8 business units / 20+ teams at 95%+ platform accuracy; drove launches with shared success metrics, positioning, and adoption playbooks. • Owned ~$7M annual budget and LLM/vendor strategy - build-vs-buy across 3 vendors + internal PoC, integrated in ~4 weeks vs a ~6-month build, avoiding ~$2M a year of internal engineering while staying model-agnostic as frontier models improve.

Product Management

UBS, Credit Suisse

Jan 2017 — Jan 2019 · Greater New York City Area

Built productivity and analytics tools for investment bankers across Credit Suisse and UBS — environments where every feature touches compliance and nothing ships without legal, risk, and regulatory sign-off. • Led a 25-person global cross-functional team (Engineering, Design, Data) that shipped iOS/web workflow tools + analytics, boosting investment-banker productivity ~30%. • Owned end-to-end delivery (discovery → roadmap → launch) across regions; translated complex regulatory requirements into product decisions that kept teams moving. The hardest part was rarely the build — it was alignment: stakeholders across regions, competing priorities, regulatory controls that couldn't bend. Built the narratives, metrics, and sequencing frameworks that made execution possible.

Product Management

Deutsche Bank

Jan 2013 — Jan 2016 · Greater New York City Area / Singapore

Led platform modernization in prime brokerage and client servicing — workflows that were slowing onboarding, inflating operational cost, and constraining growth. • Championed a $5M modernization business case — onboarded 10+ new clients, contributed to $50M in annual revenue, and reduced operational costs $500K/year through automation and workflow redesign. • Worked across product, engineering, operations, compliance, risk, and legal to turn a fragmented servicing experience into a scalable platform.