Draft mode: dashed boxes and yellow tags mark photos and links still to add. Set DRAFT_MODE to false in config.js before going live.
Mura Lakshmanan

Invention and Innovation

An enterprise real-time data platform at Synchrony Financial (2020–2026), and the research it produced.

10B+API data requests per month
~20 msaverage response time
~7,000requests per second at peak
100+ TBhistorical statement data
50+global team
2020–2026in production, continuously refined

The problem

Before 2020, frequently requested customer and account data was retrieved through external third-party calls. Each call carried a transaction cost and added latency across customer- and client-facing channels, at the scale of Synchrony's 70.7M active accounts[1] and 100M+ consumer credit accounts[2].

What I built

No equivalent capability existed in the organization. I conceived and led a new, reusable distributed caching and real-time data architecture, not a reimplementation of an existing system, and have owned its design and evolution since 2020. I lead the 50+ person global team that builds and operates it.

How the data path changed

Served from memoryCall leaves the platformCache warming
0requests in this view
0calls leaving the platform
—served from memory

Illustration only. Proportions are indicative, not live production data.

Architecture capabilities

  • Distributed caching (GemFire) and event streaming (Kafka)
  • Cache warming to pre-position frequently requested data
  • Invalidation and consistency controls aligned to the source of record
  • Cross-foundation replication for resilience and horizontal scale
  • Configurable per consuming application or client, without redesign
  • Hybrid deployment: on-premises, private cloud and public cloud
  • Historical-statement extension: 100+ TB, 12–18 months of history, sub-100 ms response

Business impact

  • 10B+ API data requests served per month at ~20 ms average
  • $12–15 million annual cost avoidance from reduced external calls, changing how the business funds and operates its digital infrastructure
  • Underpins web, mobile, customer-service and virtual-assistant experiences
  • Powers analytics for customers, external clients and internal teams
  • No re-platforming required across six years of growth (2020–2026)

Cost avoidance is my own figure from internal records; no public source exists.

Enterprise context

Independently verifiable

  • Synchrony: 70.7M active accounts as of 31 Dec 2025[1]
  • Synchrony: 100M+ consumer credit accounts[2]
  • Walmart program via OnePay, with Synchrony as exclusive issuer (June 2025)[3]
  • Amazon and PayPal named as e-commerce partners[4]
  • Amazon relationship renewed, Q2 2025[5]

Also responsible for

Application architect for Synchrony's rewards platform, which supports the loyalty programs Synchrony funds for cardholders: points, cash back, statement credits and merchandise discounts across private label, Dual Card and co-branded cards. For a sense of scale, Synchrony's loyalty program costs grew by $343 million, or 52.9%, in 2021 alone[9]. That work informs the loyalty research below.

Research built on this work

Each paper formalizes a problem I first met in production. Newest first.

  1. Presenting at CAIS 2026Add photoimages/cais-2026.jpg
    Session photo, presenter slide or certificate

    Multi-Agent Orchestration for Agentic Commerce: Coordinating Shopping, Payment, Rewards, and Fraud Agents in Real-Time Financial Transactions

    CAIS 2026 · International Conference on Artificial Intelligence and Simulation, Kuala Lumpur · Presented 29 September 2026

    How shopping, payment, rewards and fraud agents can coordinate within a single real-time transaction, built on the same kind of low-latency data infrastructure this platform provides.

    ConferenceNeeded: paper link

  2. AI-Augmented Loyalty: Integrating LLMs and Agentic AI into Rewards Platforms for Personalization at Scale

    International Journal of Computer Applications (IJCA) · Vol. 187, No. 136, pp. 16–21 · August 2026[7]

    Applying LLMs and agentic AI to personalization on rewards platforms, drawing on my work as architect of Synchrony's rewards platform.

    Read the paperPDFDOI 10.5120/ijca97298325f6eb
  3. Title slide, ECPDC 2026Add photoimages/ecpdc-2026.jpg
    Title slide
    Certificate of participation, oral presentationAdd photoimages/ecpdc-2026-certificate.jpg
    Presentation certificate

    A Self-Optimizing Multi-Agent AI Caching Architecture for High-Throughput Financial Systems

    ECPDC 2026 · 3rd International Conference on Edge Computing, Parallel and Distributed Computing, Singapore · ACM ICPS proceedings · Presented 8 August 2026

    A self-optimizing, multi-agent caching architecture in which cooperating agents predict demand, prewarm keys and tune eviction, shaped directly by lessons from operating this platform.

    ConferenceCertificate (PDF)Needed: DOI / ACM Digital Library link

Proprietary implementation details, internal identifiers and system topology are intentionally not published.

References

  1. Synchrony Financial — company information (70.7M active accounts, 31 Dec 2025)
    https://investors.synchrony.com/company-information
  2. Synchrony Financial — fact sheet (100M+ consumer credit accounts)
    https://www.synchrony.com/s/fact-sheet.html
  3. Synchrony newsroom — OnePay and Synchrony to launch new credit card (June 2025)
    https://www.synchrony.com/contenthub/newsroom/onepay-and-synchrony-to-launch-new-industry-leading-credit-card.html
  4. Synchrony Financial Form 10-K, 2018 — Amazon and PayPal named as partners
    https://www.sec.gov/Archives/edgar/data/1601712/000160171219000053/synchrony1231201810k.htm
  5. The Motley Fool — Synchrony Q2 2025 results, Amazon relationship renewed
    https://www.fool.com/data-news/2025/07/24/synchrony-financial-q2-earnings-rise/
  6. ECPDC 2026 — 3rd International Conference on Edge Computing, Parallel and Distributed Computing
    https://www.ic-ecpdc.org/
  7. Lakshmanan, M. AI-Augmented Loyalty. International Journal of Computer Applications 187(136), 16–21, Aug 2026. DOI 10.5120/ijca97298325f6eb
    https://www.ijcaonline.org/archives/volume187/number136/ai-augmented-loyalty-integrating-llms-and-agentic-ai-into-rewards-platforms-for-personalization-at-scale/
  8. CAIS 2026 — International Conference on Artificial Intelligence and Simulation, Kuala Lumpur, 28–30 Sept 2026
    https://www.ic-ais.net
  9. Synchrony Financial Form 10-K, FY2021 — loyalty programs cost increased $343 million (52.9%)
    https://www.sec.gov/Archives/edgar/data/1601712/000160171222000053/syf-20211231.htm
  • Link needed: ECPDC 2026 paper DOI / ACM Digital Library link
  • Link needed: CAIS 2026 paper or proceedings link

Talks, reviews, collaborations.

[email protected]LinkedInORCID