ARUP KANTI DAS

Solutions Architect with over 9 years of experience in post-sales support, specializing in the development and deployment of B2B SaaS platforms. Expertise includes ETL integrations, data migrations, and AWS-based backend systems, consistently delivering outcome-driven results. Successfully managed 150 tenant migrations, reducing incidents from 20 to 2 and disputes from 50 per week to 10 per week through innovative offline-first punching and near real-time commerce synchronization via webhooks. Currently pursuing a PGDM in Product & Strategy at MDI Gurgaon to enhance strategic acumen and drive cost-to-serve improvements through right-sizing and scheduled scaling initiatives.

• Binary Recognition Award, Thoughts2Binary — for leadership and bridging engineering with business outcomes.
• Supervised team of 15 staff members.
• Resolved product issue through consumer testing.

HackerRank Rest API (Intermediate) Certificate , HackerRank

• Binary Recognition Award, Thoughts2Binary, for leadership and bridging engineering with business outcomes.
• HackerRank Rest API (Intermediate) Certificate, HackerRank, 2021-04-01

• Remote
• Relocate
• Hybrid
• Travel
• On-Site

Jan 2023 - Jan 2024Jan 2023 - Jan 2024Webhook-driven Commerce Event Pipeline (Polling → Webhooks)Webhook-driven Commerce Event Pipeline (Polling → Webhooks)

• Problem: 5-minute polling caused stale order/shipment updates and wasted compute during low activity hours.
  
  Approach: Replaced polling with webhooks and added durability + correctness (DLQ/replay, idempotency/dedupe, ordering controls, signature validation) plus per-tenant throttling/backpressure.
  
  Outcome: Improved freshness from ~4+ min lag → near real-time; reduced Lambda cost ~15%; prevented duplicate/out-of-order state updates.
  
  Stack: Webhooks, API Gateway, queues/DLQ, object storage for replay, reconciliation jobs, CloudWatch/metrics.Problem: 5-minute polling caused stale order/shipment updates and wasted compute during low activity hours. Approach: Replaced polling with webhooks and added durability + correctness (DLQ/replay, idempotency/dedupe, ordering controls, signature validation) plus per-tenant throttling/backpressure. Outcome: Improved freshness from ~4+ min lag → near real-time; reduced Lambda cost ~15%; prevented duplicate/out-of-order state updates. Stack: Webhooks, API Gateway, queues/DLQ, object storage for replay, reconciliation jobs, CloudWatch/metrics.
• Skills: Webhooks · API Gateways · Message Queue · Observability

Jan 2018 - Jan 2019Jan 2018 - Jan 2019Wave-based Multi-tenant Migration Program (150 tenants, 10M rows)Wave-based Multi-tenant Migration Program (150 tenants, 10M rows)

• Problem: Modernize a multi-tenant B2B SaaS platform without disruptive cutovers or data integrity issues.
  
  Approach: Ran tenant-scoped migration waves with FK-safe dependency ordering (masters→dependents), bulk loads, and automated reconciliation to keep legacy + new systems usable during transition.
  
  Outcome: Migrated 10M rows across 150 tenants in 20 waves with negligible weekend cutover impact; reduced migration incidents 20 → 2 over the rollout.
  
  Stack: SQL bulk inserts, batch workers, migration scripts, monitoring/alerts, automated reconciliation.Problem: Modernize a multi-tenant B2B SaaS platform without disruptive cutovers or data integrity issues. Approach: Ran tenant-scoped migration waves with FK-safe dependency ordering (masters→dependents), bulk loads, and automated reconciliation to keep legacy + new systems usable during transition. Outcome: Migrated 10M rows across 150 tenants in 20 waves with negligible weekend cutover impact; reduced migration incidents 20 → 2 over the rollout. Stack: SQL bulk inserts, batch workers, migration scripts, monitoring/alerts, automated reconciliation.
• Skills: SQL · Proactive Monitoring · Data Migration

Jan 2017 - Jan 2018Jan 2017 - Jan 2018Offline-first Kiosk + Mobile Attendance with Face VerificationOffline-first Kiosk + Mobile Attendance with Face Verification

• Problem: Low-connectivity sites and shared kiosks needed reliable punching with stronger integrity and fewer payroll disputes.
  
  Approach: Built a unified Punch API across kiosk + mobile, implemented encrypted offline event queue with 24-hour offline operation and idempotent sync; integrated AWS Rekognition server-side to accelerate go-live.
  
  Outcome: Reduced disputes 50/week → 10/week (80% down); stabilized face match pass rate at ~85%.
  
  Stack: AWS Rekognition, offline queue + batch sync, device enrollment/auth, idempotent ingestion, observability.Problem: Low-connectivity sites and shared kiosks needed reliable punching with stronger integrity and fewer payroll disputes. Approach: Built a unified Punch API across kiosk + mobile, implemented encrypted offline event queue with 24-hour offline operation and idempotent sync; integrated AWS Rekognition server-side to accelerate go-live. Outcome: Reduced disputes 50/week → 10/week (80% down); stabilized face match pass rate at ~85%. Stack: AWS Rekognition, offline queue + batch sync, device enrollment/auth, idempotent ingestion, observability.
• Skills: AWS Rekognition

AI Shopping Assistant (Magento + Bedrock Agent + RAG + Actions)AI Shopping Assistant (Magento + Bedrock Agent + RAG + Actions)

• Problem: Accessibility-first Magento marketplace needed faster, guided product discovery with strong privacy and audit expectations (NDIS context).
  
  Approach: Built a Bedrock Agent backed by a catalog RAG knowledge base (vector KB) and integrated Magento actions via Lambda action groups for grounded search, comparisons, and add-to-cart flows; added safety/grounding rules to prevent unsourced catalog claims.
  
  Outcome: Enabled users to reach a relevant shortlist faster via chat-led discovery; designed tracking for ≤2-turn shortlist rate, chat-to-cart rate, and unsourced-claim rate (target near-zero).
  
  Stack: Magento 2 (REST/GraphQL), AWS Bedrock Agents + Knowledge Base, Lambda action groups, retrieval + reranking, CloudWatch/observability, privacy controls.Problem: Accessibility-first Magento marketplace needed faster, guided product discovery with strong privacy and audit expectations (NDIS context). Approach: Built a Bedrock Agent backed by a catalog RAG knowledge base (vector KB) and integrated Magento actions via Lambda action groups for grounded search, comparisons, and add-to-cart flows; added safety/grounding rules to prevent unsourced catalog claims. Outcome: Enabled users to reach a relevant shortlist faster via chat-led discovery; designed tracking for ≤2-turn shortlist rate, chat-to-cart rate, and unsourced-claim rate (target near-zero). Stack: Magento 2 (REST/GraphQL), AWS Bedrock Agents + Knowledge Base, Lambda action groups, retrieval + reranking, CloudWatch/observability, privacy controls.
• Skills: Magento · AWS Bedrock · AWS Lambda

• https://www.linkedin.com/pulse/perceived-performance-business-problem-just-latency-arup-kanti-das-zm1ac
• https://www.linkedin.com/pulse/perceived-performance-business-problem-just-latency-arup-kanti-das-zm1ac

• Binary Recognition Award, Thoughts2Binary, for leadership and bridging engineering with business outcomes.
• HackerRank Rest API (Intermediate) Certificate, HackerRank, 2021-04-01