Preeti Agrawal

Post layoffs at AlphaSense, wanted hands-on experience in end-to-end business ownership to validate AI-driven solutions in a real market environment.

Led complete business lifecycle—sourcing, vendor negotiations, pricing, inventory planning, fulfillment, and customer delivery for an AI-enabled D2C fashion brand.

Improved supplier economics through strategic negotiations. Implemented inventory discipline practices. Optimized fulfillment workflows for efficiency.

Vendors & suppliers, fulfillment partners, payment providers (Razorpay), logistics (Shiprocket), platform (Shopify).

Fashion retail suffers from high deadstock rates due to trend volatility. Traditional buying decisions rely on intuition rather than data signals.

Prototype demand-sensing system correlating real-time fashion trends with historical sales to enable faster, data-driven assortment decisions.

Multi-signal correlation: social trends (Instagram, Pinterest), search intent (Google Trends), inventory levels, and cultural events. Pattern DNA logic for trend scoring.

Built independently using Gemini Vision, Apify scrapers, Supabase, and n8n orchestration.

D2C brands struggle with high CAC and low retention. Industry benchmarks for cart recovery are 8-10%.

Comprehensive growth system covering acquisition channels, abandoned cart recovery, and customer reactivation campaigns.

Built data-driven strategies across paid ads, email sequences, and social engagement. Segmented users by behavior for personalized outreach.

Implemented using Shopify, email automation tools, Meta Ads, and analytics dashboards.

Manual content creation and lifecycle messaging is time-consuming for solo founders. Needed to scale output without scaling team.

GenAI-powered automation for email sequences, product descriptions, social content, and customer communications.

Applied generative AI (GPT, Claude) with custom prompts for brand voice. Built n8n workflows for automated content pipelines.

Self-built using LLM APIs, n8n automation, and integration with Shopify and email platforms.

Customers lacked visibility into how value was being realized. CS teams relied on manual data pulls. Renewal and expansion conversations were qualitative, not data-backed.

End-to-end client-facing analytics platform exposing feature usage, engagement metrics, and actionable insights for Admins, Power Users, and Leadership.

Partnered with CS/Sales to map decision journeys. Designed "Insight Cards" to convert data → recommendations. Defined metric definitions aligned to customer outcomes.

Customer Success & Sales (discovery), Data Engineering (instrumentation), Frontend & Backend Engineers, Leadership stakeholders.

Legacy system was batch-driven, slow, and failed at scale. Data leakage and missing datapoints created low trust. ~3M events/day with ~30% MoM growth made reliability a business risk.

Led product transition to real-time, event-driven analytics supporting high-volume ingestion, data quality, and near real-time decision-making.

Defined requirements for event ingestion, validation & schema consistency. Introduced data contracts. Balanced latency vs accuracy vs scalability. Standardized KPIs.

Data Engineering, Platform & Infra teams, Product & Engineering consumers, Leadership for migration planning.

Teams repeatedly depended on analysts for SQL queries, metric clarifications, and recurring questions. This slowed decisions and overloaded the analytics team.

Conversational BI chatbot enabling non-technical teams to query data using natural language with Dialogflow integration.

Identified high-frequency queries. Designed conversational flows, fallbacks, and clarification states. Improved intent accuracy from ~72% → ~88%.

Analytics team, Data Engineering, Backend Engineers, Business stakeholders.

Existing NPS setup had low response rates, poor segmentation, and limited actionability for product decisions.

Redesigned NPS collection and analytics infrastructure to improve signal quality and downstream usability.

Duplicate SKUs degraded CX, fragmented reviews, and created merchandising inefficiencies. Manual review was slow, error-prone, and couldn't scale.

End-to-end automated deduplication using NLP + Computer Vision with confidence-based automation and human-in-the-loop for low-confidence cases.

Combined image similarity (CNN embeddings), text similarity (cosine), and confidence scoring. Built review workflow for approve/reject/merge decisions.

ML Engineers, Catalog Operations, Platform Engineers, Business stakeholders.

Significant SKUs were delisted due to catalog defects, still commercially viable, but invisible to customers due to manual resolution delays → lost sales and poor seller experience.

Automated defect-resolution workflows that identify eligible SKUs, resolve common defect patterns, and restore buyability at scale.

Identified high-frequency defect categories. Designed 3 automated workflows: detection → validation → resolution. Built safeguards for incorrect reinstatements.

Catalog Ops teams, Platform & tooling engineers, Seller experience stakeholders.

Physical lookbook campaigns had high printing/distribution costs, risked waste from duplicate recipients, and lacked precise household-level targeting.

A/B testing framework with household-level deduplication to measure incremental lift and optimize campaign reach and ROI.

Built deduplication algorithm for household-level targeting. Created test vs control cohorts. Tracked incremental conversions and sales uplift.

Marketing teams, Data science & analytics, Campaign operations, Finance stakeholders.