Bio - A research lab that intercepts raw satellite internet radio signals using a custom rooftop antenna dish. Because space satellites move at insane speeds, their signals get messy. This lab builds software to clean up that data and track individual satellites. So, I built the entire coding pipeline from scratch to turn messy radio waves into neat, computer-labeled satellite tracks.

• Navigated extreme ambiguity and exercised autonomous ownership to architect a 0→1 data-processing platform in Python and Streamlit for proprietary Starlink downlink signals, replacing manual workflows with satellite-identified datasets used daily by the research team.
• Eliminated track-assignment collisions across overlapping Doppler-shift S-curve trajectories by implementing a two-stage pipeline using DBSCAN clustering and the Hungarian Algorithm to resolve high-density signal detections against theoretical trajectories.
• Achieved 90%+ automated detection accuracy by implementing real-time signal observability and automated regression testing queues that suppressed side-lobe leakage from 53.5 dB to 14.1 dB, cutting data turnaround time from hours to under one minute and deployed interactive analysis through Docker on GPU infrastructure.

Bio - B2B SaaS digital optimization and enterprise split-testing platform serving 3,000+ global brands. So, I ran the product line that added autonomous AI agents to the platform, building the evaluation pipelines that automatically monitored and improved these web tests and server side testing.

• Owned the production-grade AI Quality & Evaluations roadmap for Wandz, VWO's agentic AI layer, defining reliability standards, LLM groundedness benchmarks, and release guardrails across 15K–20K production agent actions/week spanning 40–50 enterprise accounts through LLM-as-judge evaluation, failure taxonomy, and continuous quality telemetry.
• Helped drive critical agentic-platform error rates from ~9% to <3.5% in two months (60%+ reduction) by introducing human-in-the-loop exception routing and AI safety validation gates that blocked incorrect context interpretation, context rot and independently verified experiment variants.
• Reduced AI evaluation turnaround from 1–2 days to <1 minute and manual QA effort by ~4–5 hours/week per evaluator and uplifted developer experience by replacing sample-based reviews with full-coverage automated scoring and a centralized evaluation application adopted daily by 4 product teams / 12–15 users.
• Redirected engineering investment from an underlying LLM upgrade to agent-harness reliability by building a benchmark that separated model capability from multi-step execution failures, enabling product teams to prioritize fixes against the failure modes affecting production customers.
• Captured $10M in incremental annualized revenue for enterprise e-commerce merchants by steering the rollout of server-side feature experimentation capabilities; managed the execution of 50+ Server-Side tests utilizing VWO's Bayesian SmartStats engine to verify pricing relevance and lift checkout conversion rates by 4.2%

Bio - B2C healthcare marketplace specializing in Medicare insurance enrollment. Healthcare laws are incredibly confusing, causing seniors to frequently get frustrated, make mistakes, or drop out of the checkout funnel. So, I designed the smart, step-by-step sign-up flow to prevent compliance errors and make picking a doctor simple.

• Generated 300+ net-new digital enrollments within 6 months and ahead of the Annual Enrollment Period by launching the company's first end-to-end Medicare marketplace, digitizing quote, eligibility, plan comparison, and enrollment workflows while navigating provider-network and regulatory constraints by having full autonomous ownership to launch.
• Reduced plan-selection drop-off by 32% by translating insights from 50+ customer interviews and funnel/session analytics, customer discovery loops into a personalized questionnaire, ranked plan recommendations, and agent-assisted enrollment flow that simplified complex Medicare choices.
• Reduced acquisition cost from $210 to $172 per confirmed enrollment (~18%) in a controlled 50/50 digital-first vs. agent-first cohort by using HubSpot call transcripts and Mixpanel funnel data to identify commitment concerns, then adding provider/prescription lookup and disenrollment-rights guidance to the plan-comparison flow.
• Improved Medicare quote accuracy by 25% (61% → 76%) by defining location-resolution and provider-network matching rules that prioritized verified ZIP/address data over unreliable IP detection, while partnering with engineering to surface real-time doctor network status using refreshed insurer data.
• Eliminated compliance violations flagged in internal reviews across 12 states over 9 months by translating CMS, HIPAA, and state-specific eligibility requirements into ~40 reusable validation scenarios, guardrails, contraints, and instructions that automatically gated enrollment paths and routed ambiguous cases to licensed agents.

Stallion Express is a high-growth B2B/B2C cross-border e-commerce logistics and shipping aggregator platform processing 5M+ annual parcels for 40,000+ active digital merchants. Managed the technical product roadmap and GTM strategy for shipping discovery, geospatial search, and multi-currency checkouts..

• Reduced checkout abandonment by 14% in an A/B test by launching shipping discovery that showed customers multiple carrier options with transparent 70%+ savings comparisons, reducing uncertainty around shipping costs at checkout for a platform serving 40,000+ merchants.
• Spearheaded a 0-to-1 MVP proposal targeting a $9M cost-reduction opportunity for a native self-service parcel drop-off and label-management workflow across iOS and Android.
• Increased first-shipment activation by 22% by introducing location-based personalization that identified a new user's nearest shipping hub and surfaced relevant shipping options without requiring manual location entry. and supported a total annual shipping volume of 5M+ processed parcels.
• Enabled 10+ new Drop & Go locations across Montreal, Vancouver, and Surrey, supporting 32 regional pricing zones and $1.2M in incremental GMV, by defining requirements for location-specific carrier pricing and partnering with Engineering to replace a single global rate model with reusable regional pricing.
• Minimized payment friction and cross-border FX drop-offs by leading the technical integration of Stripe and PayPal multi-currency payment gateways; processed $2M+ in monthly transactional credits, allowing US and Canadian merchants to pay in native currency while eliminating a 2.5% foreign exchange transaction penalty.

Cloud PaaS automation firm specializing in infrastructure orchestration. So, I led the product team that designed a self-serve interface from scratch, shrinking a technical 8-hour manual setup down to a 15-minute automated click.

• Led Engineering and UX/UI teams to build and launch a 0→1 shared services framework and AWS PaaS product in 12 months, enabling enterprise customers to create cloud infrastructure applications in <4 hours through a guided self-service experience.
• Defined product vision, KPIs, PRDs, pricing, and roadmap through market research, AWS/GCP competitive analysis, customer interviews, and private-preview usage data, focusing the MVP on four AWS service types that covered ~80% of customer demand.
• Grew active customer adoption 20x, from 5 private-preview design partners to ~100 customers, by prioritizing the highest-value features, removing onboarding friction, and expanding self-service adoption through the AWS Marketplace and customer-led growth.
• Reduced 3-tier AWS deployment time by >95%, from 6–8 hours to 15–20 minutes, by redesigning the guided provisioning experience to automate resource sequencing and configuration.
• Unblocked ~75% of early-access provisioning attempts by identifying resource naming as a major onboarding failure, then prioritizing real-time validation and suggestions before customers reached the end of the setup flow.
• Reduced custom engineering effort for partner integrations by defining reusable API requirements and integration patterns for 1P and 3P integrations with Engineering.

Cloud-native enterprise resource planning (ERP) platform acting as the central operating system for hotels.. Their software runs the entire hotel backend- handling guest room assignments, housekeeping tasks, credit card bills, and vendor supply ordering. So, I was the product manager intern who reshaped their vendor order-management flows, making it much faster for managers to restock hotel inventory.

• Increased customer onboarding by 25% in one quarter by shaping product strategy across core ERP workflows including purchasing, inventory, and payments, and prioritizing new features based on customer needs.
• Increased monthly active users by 30% in three months by simplifying customer onboarding through a vendor order-management workflow that consolidated key purchasing activities.
• Increased average monthly meetings by ~10% by launching a self-service "Create Meeting" capability in the web application.
• Established product planning and delivery processes by translating UX research into OKRs, epics, user stories, and Jira-ready requirements for Engineering.