Ravi Singhal

Shaping the target architecture and multi-year direction for a commercial banking AI assistant platform — how agent capability scales without outpacing auditability, oversight and the bank's risk appetite.

Researching and prototyping the agent identity model for the tool plane — SPIFFE workload identity, OAuth 2.1 delegated token exchange so an agent acts on behalf of a named user, and an effect ceiling of min(agent, human) at the gateway.

Architecture for API monetization and the Open Banking Gateway: metering and entitlement for paid API products; consent and authorisation lifecycle under Section 1033 and FDX.

Building the platform that powers ServiceNow's AI-native product roadmap, cloud-native, hosted on AWS EKS and Bedrock

GenAI / LLM platform on Kubernetes. Built an Ollama-based model serving layer, LangChain-powered RAG pipelines grounded in Oracle DB, and a full observability stack covering trace collection, automated evaluation pipelines, drift detection, and model reliability monitoring.

Oracle Verrazzano (GA August 2021), Oracle's open-source enterprise multi-cluster Kubernetes platform. Designed multi-cluster container deployment capabilities, built the Go operator-driven control plane, and engineered the full observability stack (Prometheus, Grafana, Jaeger, Kiali, OpenSearch, Istio, Rancher) automated via declarative CRDs. Delivered the ReactJS management console.

Helped secure Verrazzano's first enterprise paying customer, a major Japanese enterprise, by owning the coordination across their teams and ours. It was the platform's first revenue, and it shaped the product roadmap.

India engineering team for OCI CloudShell, Oracle Cloud's in-browser developer terminal, bringing secure, browser-native CLI access to Oracle Cloud users globally. Earlier roles included Wercker container-native CI/CD and WebLogic Server platform engineering.

Owned the prototype and evaluation of the bank's first streaming ML pipeline for the Digital Frontline team, social data ingestion via Kafka and Storm, LingPipe model inference — and the recommendation that came out of it on what was worth productionising.

Owned the systems architecture for a zero-downtime omnichannel platform serving 20+ million retail banking users across mobile, web, and ATM, designing for continuous availability, cross-channel consistency, and the latency requirements of transactional banking at UK scale.

Built the platform on Java Spring Boot microservices hosted on Azure, with Flume and Solr behind them for log analytics, sub-second search across high-throughput production workloads in global banking infrastructure.

Served as the technical translation layer between senior UK and India leadership, converting executive strategy into detailed engineering roadmaps, driving cross-functional alignment, and representing platform architecture in executive reviews spanning two geographies.

Built distributed Hadoop/MapReduce/Sqoop data processing pipelines inside Wellington Management, enabling investment research teams to run quantitative parameter computations across massive financial datasets that were previously too slow or too large to process interactively.

Engineered a high-performance real-time quantitative pricing engine using Thrift-based distributed RPC: live market data streams in, options pricing model execution, and sub-millisecond results published to a Memcached caching layer consumed by downstream analytical models. Designed for the throughput and latency demands of active trading environments.

Rearchitected a Java and PL/SQL data extraction pipeline that had become a bottleneck, reducing the time to generate 16+ million record Tableau-ready extracts from a full working day to under 2 hours, giving near-real-time business intelligence to investment decision-makers.

Took on a full rewrite of RBS Global Trade's legacy Visual Basic BI and charting platform, migrating it to a modern GWT, Java, JPA, Hibernate, JPivot, and Mondrian stack. The result: eliminated brittle legacy dependencies, gave the Global Trade division a maintainable analytics foundation, and made it possible to extend reporting that the old technology had frozen.

Delivered OSS integrations for AT&T BusinessDirect, including Cramer SSO and EJB portal metrics, enabling faster, more reliable telecom service activation workflows across AT&T's enterprise carrier infrastructure.

Built Telecom BSS integrations (JAX-WS web services, CORBA/JMS) for digital ad space contract billing, connecting disparate BSS systems to automate billing workflows that had previously required manual reconciliation.

Built Java/XForms tooling for the Open Text self-service BPM platform — reducing development time for key application workflows by 80% and enabling teams to ship BPM applications in a fraction of the previous effort.

Developed a Struts/MySQL bug tracking system that improved defect tracking efficiency by 70%, bringing structure and visibility to a previously ad-hoc process in a large global engineering environment.