A panelist & speaker session from London-based engineer Mohammed Nasiruddin walked the South Asian Oracle developer community through wiring OCI Generative AI, OpenAI and Cohere into Oracle APEX Low Code Platform workflows to build and showcase AI-augmented applications with enterprise integration patterns.
Mohammed Nasiruddin (panelist, far right) with fellow speakers at Oracle APEX Utsav 2024, Hyderabad, 19 October 2024; and (right) presenting the Generative AI and Fusion Integration session. Photo: INOAUG.
HYDERABAD – Oracle APEX Utsav 2024, the principal community gathering for the Oracle APEX developer community across South Asia, drew several hundred practitioners to Hyderabad on 19 October for a programme of sessions centred on a single, increasingly urgent enterprise question: how to add generative AI capability to a long-established low-code platform without re-platforming the systems that sit underneath it.
Among the speakers on the day was Mohammed Nasiruddin, a London-based platform engineer whose session, titled “Unlock exclusive insights on Generative AI and Fusion Integration with Oracle APEX,” walked attendees through a practical pattern set for connecting OCI Generative AI, OpenAI and Cohere as providers inside the Oracle APEX low-code environment. The talk covered authentication, REST data source design, and integration patterns specifically aimed at workflows already running against Oracle Fusion application data in production.
What the session actually covered
Where many AI-on-low-code talks in the past year have stopped at general-purpose demonstrations, Nasiruddin’s session was pitched at engineers and architects already maintaining APEX estates inside large enterprises. The walkthrough covered configuring OCI Generative AI, OpenAI and Cohere as Generative AI Services inside Workspace Utilities; the four authentication patterns most commonly required to wire AI providers into APEX REST Data Sources (Basic, OAuth 2.0 Client Credentials, HTTP Header credentials, and URL Query String); and integration patterns that allow document-driven and form-driven workflows to be augmented with AI without breaking the underlying transactional model.
He also stepped attendees through implementation patterns for the in-built APEX Assistant against Oracle Fusion datasets at production scale. The angle, attendees said afterwards, was less about novelty and more about reproducibility: a working pattern that an enterprise APEX team could adopt without commissioning new infrastructure.
Panelist Q&A: Audience questions on APEX AI Assistant with OpenAI and Cohere
Following the main session, Nasiruddin joined the event panel and fielded a large volume of practitioner questions from the floor. A significant portion centred on the Oracle APEX AI Assistant and its behaviour when configured against OpenAI and Cohere as external providers. The following is an edited record of the exchanges.
Q: When we configure the APEX AI Assistant to use OpenAI rather than OCI Generative AI, the Assistant loses schema awareness and no longer understands our application data model. How do you handle that?
A (Nasiruddin): That is the most common friction point, and it is architectural rather than a configuration error. The APEX AI Assistant’s schema-awareness is tightly coupled to OCI Generative AI because Oracle controls both sides of that integration. When you route to OpenAI via a REST Data Source, you lose that metadata feed. The approach that holds in production is to pre-load a condensed schema summary — table names, key columns, business context — into the
system prompt of the OpenAI call. You are manually providing what OCI would have injected automatically. It creates a maintenance overhead when your schema changes, but it is reproducible and debuggable, which is the real test.
Q: We are evaluating Cohere Command R for our APEX estate because of its retrieval capability. Does the REST Data Source pattern support streaming responses, and does APEX render them?
A (Nasiruddin): Cohere Command R is a strong choice where RAG is the primary pattern — its grounding and citation format maps well to structured APEX display components. On streaming: APEX’s native REST Data Source is a request-response model and does not handle server-sent events natively. To stream you would need to manage the stream at the PL/SQL layer via UTL_HTTP or a thin middleware shim. For most ERP-adjacent workflows this is unnecessary — the latency on a well-scoped Cohere call against structured data is low enough that non-streaming
serves users adequately. Invest in streaming only if you have long-form generative output in the UI, which is uncommon in back-office workflows.
Oracle APEX Utsav is run by INOAUG, the regional Oracle user group, and is one of the more influential APEX-developer venues globally by community reach. Speaker invitations are issued on technical merit by the user community itself, rather than by sponsorship, which makes the speaker roster a credible signal of practitioner standing rather than a paid stage.
The event draws engineers and architects from enterprise back-office estates across South Asia and beyond, including teams working on ERP, finance, HR, supply chain and regulated industry deployments where Oracle APEX has long served as the low-code layer above large Oracle Database installations.
The UK angle
Oracle APEX is in active use across UK enterprise back-office estates, including NHS supply chain functions, public-sector ERP deployments, and regulated financial services. The pattern set Nasiruddin demonstrated in Hyderabad maps directly onto a challenge UK enterprise estates are now expected to address under the Government’s AI Opportunities Action Plan, written by Matt Clifford and published in January 2025, which the British government has accepted in full. The Plan has been credited with catalysing approximately 14 billion pounds in private-sector AI commitments and around 13,000 new AI roles in its first wave.
For an APEX-anchored enterprise estate that cannot afford the timeline or cost of a full re-platforming exercise, the ability to add AI capability inside the existing low-code shell, with familiar authentication and REST integration patterns, is the difference between an AI roadmap that ships in months and one that disappears into a multi-year migration. The Department for Science, Innovation and Technology has placed the UK AI sector at more than 5,800 companies generating around 23.9 billion pounds in annual revenue, and the operational question of how non-frontier enterprise estates participate in that growth is one of the more practical themes of the next two years.
TechBullion reached Nasiruddin for comment after the session. He declined to characterise the work as novel and pointed instead to the practitioner audience as the test. “The interesting question is not whether you can wire a model into APEX,” he said. “It is whether the engineers maintaining the system at three in the morning can debug it, hand it over, and trust it. That is what these patterns are for.”
About the speaker
Nasiruddin is a London-based product and platform engineer with 16 years of experience across the UK, Gulf and India. He serves on the platform engineering team at The Guardian, supporting infrastructure for over 100 million monthly readers, and is technical co-founder of Matchpoint AI, a London venture connecting investors with capital-seeking businesses. His prior delivery record spans a multi-tenant safety management platform across 15 enterprise customers, an asset management platform tracking over $50 million in assets, and an e-invoicing platform processing more than 2 million invoices per month.
The broader trend
Hyderabad’s Oracle developer community is one of the largest in the world, and sessions at INOAUG events have historically served as informal reference architectures that propagate into enterprise environments globally. Engineers at APEX Utsav 2024 noted a shift in the day’s programme: AI moving from a discrete project to a routine integration concern, sitting alongside identity, audit and observability inside the standard enterprise stack.
What comes next
Selected APEX Utsav 2024 session recordings are expected through INOAUG community channels in the weeks following the event. Nasiruddin intends to publish accompanying reference material on his public GitHub portfolio at github.com/mnk-nasir , where he maintains a continuing contribution record, and to extend the integration patterns into a longer-form write-up for the UK APEX practitioner community.
Last updated: June 2, 2026
