Haneet Singh Chhabra

I work closely with product and leadership throughout the feature lifecycle — not only during implementation. This is a core part of how I deliver as a senior Android engineer.

Examples

• Video reviews — self-initiated from discovery through 3-sprint delivery plan; partnered with PM on user stories and aligned vendor (BV) on API behavior.
• UGC 2.0 architecture — evaluated 6+ enrichment approaches across 5+ platform teams; recommended IRO + LRU Cache and briefed senior leadership.
• Creator vs Reviewer Program — identified a product constraint no team owned; designed APIs and aligned both program teams on enforcement.
• Used & refurbished products — Android technical lead coordinating requirements and timelines across backend, web, and product.
• AI image search — partnered with data science and search teams from prototype scoping through Android delivery.

Walmart Shopping App · Reviews & User-Generated Content

Android Performance & Architecture

• Owned reviews UI performance on the item detail page — render time 380ms → 96ms, CPU −72%, dropped frames −73%, jank eliminated on one of Walmart's most-visited Android screens.
• Led variant reviews caching on Android — reduced 1–1.6 billion API calls per day and improved latency ~21% without business impact.
• Migrated reviews from a legacy embedded module to a standalone Android module — zero production regressions, full backward compatibility.
• Architected parallel dashboard data loading on Android with timeout controls, partial-failure handling, and graceful degradation.

Android Features — Owned E2E

• Camera upload and multi-image selection — presigned URL flows, image compression, and direct-to-cloud upload integration on Android.
• Reviewer community and direct camera entry — owned from PM discovery through Android implementation and iOS/web alignment.
• Video upload pipeline — shipped streaming upload with 64KB chunked RequestBody, phase-based progress state machine (IDLE → UNDERSTANDING → UPLOADING → PROCESSING), and O(1) moov-atom metadata extraction in 20ms. BV integration with Passkey auth.
• GenAI image search prototype — partnered with data science and search teams; vector similarity and on-device model integration on Android (<2s latency).
• Android technical lead for used/refurbished product experiences — coordinated delivery across backend, web, and product teams.

Backend & API (When Features Require It)

• Contributed to GraphQL API orchestration for reviews platform rebuild — resolvers, seller reviews, product sampling, and catalog enrichment.
• Built OpenAPI service clients and integration patterns adopted team-wide; designed Redis caching projected to cut peak backend load ~85%.
• Implemented memcache layer for IRO seller reviews — cut P95 latency ~40% for online seller reviews API.
• Led UGC 2.0 item enrichment architecture investigation — evaluated 6+ approaches (IQS, P13N, Preso/Search, Item Service, GraphQL Federation, IRO) across 5+ platform teams; recommended IRO + LRU Cache strategy.

Quality, Tooling & Leadership

• Built PixelMind (DINOv2-based visual regression testing) and deployed to WCNP/Kubernetes — Docker image to Walmart container registry, Kubernetes seed Job to copy DINOv2 model weights from Azure Blob to PVC for air-gapped deployment. Adopted as team standard.
• Built Automaton cron-based scheduled performance testing — Looper → curl → Automaton pipeline for unattended REST API perf runs; 30–35% reduction in automation false failures via skip functionality.
• Sustained 48+ consecutive stable CI runs; led incident response when automation dropped to ~48%.
• Designed A/B experiments with clear success metrics; mentor 2–3 Android engineers on performance and architecture.

UGC Pipeline Jobs — Large-scale ETL, NLP moderation, and ML classification for product reviews & Q&A

Spark ETL & Data Lake Architecture

• Built multi-module Apache Spark 3.4 pipeline system processing millions of product reviews and Q&A — ingestion, transformation, and loading into Apache Hudi data lake, Elasticsearch, and Cassandra.
• Designed AbstractUGCJob base class centralizing Spark/Hudi configuration, Kafka producers, SFTP upload, and multi-store data loaders across 280+ classes.
• Implemented AES-GCM encryption, PII filtering, and CCPA compliance workflows for customer data protection at scale.

NLP Content Moderation Engine

• Built rule-engine-backed content moderation filters — profanity, spam, PII, fake review, foreign language, URL, and DBA detection — using Stanford CoreNLP, Apache Lucene, and DFA automata.
• Designed pluggable filter architecture processing reviews through a moderation pipeline before publishing.

ML Model Training & Prediction

• Developed Spark MLlib classification models (Logistic Regression, Naive Bayes) with cross-validation for automated review, question, and answer categorization.
• Built prediction service applying trained models at runtime for content classification at scale.

Customer enrollment lifecycle across Promo Engine, Customer Insights, and Creators Program

• Designed review sampling enrollment system orchestrating across Promo Engine and Customer Insights services with compensating-transaction (saga) rollback on partial failures.
• Built Spring Data Cassandra repositories with composite partition keys for enrollment, history, and reminder tracking across tenants.
• Implemented HMAC signature-based API authentication, feature-flag-driven branching, and Creators Program mutual-exclusion enforcement.
• Managed full enrollment lifecycle: invite, accept, opt-out, disenroll, verified-purchase checks, and email/reddot reminder notifications.

60+ production DAGs for review syndication, purchase history, email notifications, and operational reporting

• Orchestrated 60+ production Airflow DAGs on Astronomer runtime — Spark jobs via GCP Dataproc, analytical queries in BigQuery, and object storage on GCS.
• Built pipelines for Bazaarvoice review syndication, customer purchase history, Kafka data sync, Elasticsearch ingestion, and customer segmentation.
• Implemented data-quality safeguards: freshness checks, staleness detection, staging tables, and EDW joins in templated BigQuery SQL.
• Designed observability layer with Slack alerting, xMatters escalation, and AST-based DAG-id collision tests in pytest.
• Managed CI/CD with Concord/Looper Pro, SonarQube, and Docker-based deployments.

As an Android engineer, I don't stop at the Activity layer. I drive features from first conversation to production — partnering with PM, aligning stakeholders, and writing the code myself.

Discover & Align

• Self-initiate feature investigations (video reviews, UGC 2.0 enrichment) — from vendor API research through sprint delivery plans.
• Drive architecture decisions across organizational boundaries — recently evaluated 6+ enrichment approaches spanning 5+ platform teams.

Design & Influence

• Own technical design end-to-end — module architecture, performance budgets, caching strategy, and cross-platform API contracts.
• Proactively surface blockers before downstream teams are impacted — caught API migration dependency before it reached production.

Ship & Grow the Team

• Deliver across the stack when features demand it — Android, backend services, data pipelines, and infrastructure (PixelMind on WCNP/K8s).
• Mentor 2–3 engineers; delegate design reviews; build tooling (Automaton, PixelMind) that scales team capability beyond individual output.

Next.js 15 · Neon Postgres · Vercel Blob · AWS Rekognition · Firebase · Resend

• Built a cinematic guest experience with Framer Motion, GSAP, Lenis, and React Three Fiber — invitation, countdown, story, events, venue map, gallery, and RSVP.
• Shipped a PIN + RBAC admin dashboard for events, RSVPs, accommodation/room assignment, gallery moderation, team invites, and feature flags across main / mehfil / laavan sites.
• Persisted ops data in Neon Postgres and media in Vercel Blob (direct uploads for large files on serverless); transactional email via Resend.
• Integrated AWS Rekognition for “find my photos” — selfie match, face clusters, multi-merge, and magic-link private albums for guests.

Turborepo · Next.js 14 · Firebase · Razorpay / UPI · OCR + LLM · Zod

• Designed a fintech monorepo for receipt scanning, equal/item/custom splits, secure claim links, and UPI collection with Razorpay — no platform wallet (RBI PA/PG-aware).
• Built a receipt OCR pipeline with Tesseract (default), optional Google Cloud Vision, and Gemini/OpenAI LLM parsing behind admin + per-user entitlements.
• Implemented webhook-verified payments with HMAC idempotency, money as integer paise, live collection progress, and Firestore security rules/indexes.
• Shipped admin ops: users, roles, feature flags, scan entitlements, transactions, and trends — shared packages structured for a future React Native client.