unravel / migration work

Changing core systems. Without stopping the business.

Eight verified migrations across frontend, cloud, data, SDK, and backend systems. Each case leads with business impact, elapsed duration, and complexity, then explains continuity constraints and execution choices.

Focused on judgment before mechanism.

What repeated

Judgment before machinery.

  1. Business continuity sets architecture.

    Zero downtime, customer-paced upgrades, and continuous feature work each demand different migration shapes.

  2. Parallel paths buy reversibility.

    Dual databases, coexistence, staged customer cohorts, and versioned opt-in reduce irreversible cutover risk.

  3. Contracts beat source translation.

    Consumers, wire formats, and data invariants define what must survive. Languages and frameworks are implementation choices.

  4. AI speed needs stronger gates.

    Four-week delivery was possible because human-defined contracts and live-stack checks could reject machine-speed mistakes.

The crossings.

Each entry leads with business impact, elapsed duration, and a complexity rating with reasons. Technical details follow only to explain how value and continuity were protected.

  1. 2026
    Celerity
    Java → Python backend     Returned direct product change control and reduced maintenance surface by 55%.
    Unreleased target

    Business impactRemoved vendor lock-in, returned direct product change control, reduced source maintenance surface 55%, and created 4 independently deployable and scalable processes.

    Duration4 active weeks. Original Java build took about 21 active weeks. Two more weeks would have avoided overtime.

    ComplexityHigh. Two strict mobile contracts, Kafka compatibility, hidden framework behavior, data and storage invariants, and a compressed finish.

    The change

    Java Quarkus monolith to Python FastAPI API, consumer, renderer, and dispatcher.

    The constraint

    Live iOS and Android clients defined compatibility, but Python target was unreleased. This was not a live production cutover.

    The method

    • Extract contracts from mobile clients.
    • Verify API, database, and storage behavior against real stack.
    • Use 101 client contract assertions and 247 live-stack checks per gate.

    AI-assisted?Yes. Claude Code agents did substantial build and verification work; humans defined contracts, gates, and architectural boundaries.

    Read Terra Incognita, the detailed migration account →
  2. 2025
    OneStudyTeam
    Retool → React + Clojure     $80K annual savings, with no downtime, no outage, and no loss of customizations.
    B2B healthcare SaaS

    Business impact$80K saved per year, zero downtime, and no outage with no loss of customizations. Customer UX improved while the migrated stack stayed maintainable by the engineering team.

    DurationApril through June 2025, under 3 months elapsed. 2 senior engineers focused on AI-assisted migration delivery.

    ComplexityMedium. Integrations and data contracts, relational schema with 2,000 columns, data migration, and multi-team/stakeholder coordination. Warehouse redesign was included for performance under human judgment.

    The change

    OneStudyTeam used Retool for weekly customer onboarding, updates, customization handoff, and sales/customer-success upselling across teams. The system contained more than 30,000 lines with a large custom workflow surface.

    The constraint

    The original developer had left, the system was unmaintained, customizations were frequently lost, and customer NPS dropped. A previous replacement effort exposed complex database migration and coordination across multiple teams and owners.

    The method

    • Move one Retool workflow at a time after interviews and boundary definition with users and owners.
    • Scope backend, frontend, and data changes together and redesign when needed (including warehouse performance changes).
    • Deploy every workflow to test, collect feedback, iterate, then ship to production and retire the migrated Retool workflow.

    AI-assisted?Yes. AI analyzed the 2,000-column relational schema and existing codebase, then generated synthetic data and property-based tests to fuzz data combinations.

    Human judgment2 senior engineers owned workflow sequencing, integration and data-contract boundaries, continuity gates, cross-team decisions, stakeholder alignment, and a warehouse redesign to improve performance.

    ReferenceRaman Adlakha, India Head and Director of Engineering, OneStudyTeam

  3. 2024
    HealthTech
    Angular checkout → React     Checkout conversion increased 6 to 8%, and disruption from navigation or payment windows fell 30%.
    Anonymous client

    Business impactCheckout conversion increased 6 to 8%. Disruptions caused by navigation or payment-window closure fell 30%.

    DurationAbout 3 months elapsed.

    ComplexityMedium. Revenue-critical flow, navigation and payment state risk, bounded frontend scope, and one cutover.

    The change

    Angular checkout to React, with checkout reliability improvements delivered during migration.

    The constraint

    Protect customer progress and transaction completion through refreshes, back navigation, and payment-window closure.

    The method

    • Persist early checkout state in local storage.
    • Use a reopenable payment popup.
    • Release through one controlled cutover.

    AI-assisted?Yes.

  4. 2022
    Helpshift
    Hadoop + Hive → Spark + Redshift     Supported 10x platform growth and improved data accuracy from 90% to 99.99% in one year.
    Live data platform

    Business impactSupported 10x platform growth, improved data accuracy from 90% to 99.99% in one year, and strengthened resiliency, scalability, and adoption.

    DurationAbout 6 months elapsed. The 1-year accuracy period is an outcome window, not migration duration.

    ComplexityHigh. 2 billion devices, 1 PB daily ingestion, live accuracy requirements, and a broad warehouse change.

    The change

    Hadoop and Hive to Spark and Redshift. Supporting stack included Hudi, PostgreSQL, Arrow, Airflow, and DBT.

    The constraint

    Maintain processing continuity and data correctness at petabyte daily scale.

    The method

    Modernize processing and warehouse architecture around Spark and Redshift, supported by Hudi, PostgreSQL, Arrow, Airflow, and DBT.

    AI-assisted?No.

  5. 2022
    Helpshift
    MongoDB → YugabyteDB     More consistent data operations, greater scalability, and headroom for future load.
    Zero downtime

    Business impactMore consistent data operations, greater scalability, and headroom for future load without interrupting customer service.

    DurationAbout 6 months elapsed.

    ComplexityHigh. Live 2 TB store, data integrity, dual reads and writes, and a 20-plus phase roadmap.

    The change

    MongoDB primary store to YugabyteDB distributed SQL, with supporting service extraction.

    The constraint

    Preserve data integrity and customer availability throughout cutover.

    The method

    • Build staging environment.
    • Run both databases through dual reads and writes.
    • Move gradually across more than 20 phases.

    AI-assisted?No.

  6. 2021
    2022
    Helpshift
    Native SDKs → shared JS core     Faster SDK delivery, consistent behavior across platforms, and lower duplicated maintenance cost.
    Customer-paced rollout

    Business impactFaster SDK delivery, consistent behavior across platforms, lower duplicated maintenance cost, and broader customer upgrade reach.

    Duration4 to 6 months elapsed.

    ComplexityHigh. Multi-platform compatibility, version support, customer-controlled upgrades, and coordinated releases.

    The change

    Separate iOS, Android, and Unity or game-engine SDK implementations to a shared JavaScript core.

    The constraint

    Keep existing native versions supported while customers choose their upgrade timing.

    The method

    Validate shared behavior across all three platform families, then release through versioned opt-in.

    AI-assisted?No.

  7. 2021
    Helpshift
    AWS California → Virginia     Almost $1M annual infrastructure savings, completed with zero downtime and no data corruption.
    Zero downtime

    Business impactAlmost $1M in annual infrastructure savings. Production cutover completed with zero downtime and no data corruption, while establishing a more current operating baseline.

    Duration9 months, January through October 2021, from active planning through production rollout. California decommissioning continued afterward.

    ComplexityHigh. About 1,000 instances, 120 application services, 60K requests per second, 1 PB of S3 data, 150 TB of Kylin data, and multiple state and traffic strategies.

    The change

    AWS Northern California to Northern Virginia, including state, applications, traffic, release, monitoring, and infrastructure baselines.

    The constraint

    Move a B2B customer portfolio without interruption, including a long tail and very large customers.

    The method

    • Map service dependencies, readers, writers, and data entities.
    • Establish replicated event and state paths.
    • Move traffic by customer cohort, from test domains to top accounts.

    AI-assisted?No.

    Read the full AWS migration deep dive →
  8. 2015
    2017
    Helpshift
    YUI → React     Product roadmap never paused while React improved future delivery and stability.
    Live system

    Business impactProduct roadmap never paused. React improved future delivery, performance, reliability, consistency, usability, and extensibility.

    DurationAbout 2 years elapsed.

    ComplexityHigh. Live product, broad UI surface, long coexistence, and uninterrupted feature work.

    The change

    YUI frontend to React.

    The constraint

    Keep shipping product features throughout migration.

    The method

    Move incrementally by product surface while YUI and React coexist.

    AI-assisted?No. Pre-LLM delivery.

Migration phase zero

Three weeks to find the right first move.

Discovery diagnoses the business problem before recommending architecture. It produces a phase-one decision that stands on evidence, even if another team implements it.

WEEK 01 / CONTEXT

Map what matters.

  • Interview leaders and domain owners.
  • Walk code and architecture with engineers.
  • Read telemetry, incidents, and release history.
WEEK 02 / RISK

Test the weak seams.

  • Choose spikes, operational shadowing, or data analysis per risk.
  • Verify contracts, rollback behavior, and data correctness.
  • Separate architectural symptoms from business causes.
WEEK 03 / PLAN

Make phase one executable.

  • Deliver Prioritized risk map.
  • Deliver Detailed phase-one plan.
  • Deliver Directional roadmap.

One decision: Approve phase one.

Cadence: kickoff, two weekly reviews, and final readout.

Team shape: decided after access call, based on system criticality and evidence access.

Sequence: diagnose correctly, migrate safely, then use AI where verification makes acceleration responsible.