DeepLight AI is a specialist AI and data consultancy with extensive experience implementing intelligent enterprise systems across multiple industries, with particular depth in financial services and banking. Our team combines deep expertise in data science, statistical modeling, AI/ML technologies, workflow automation, and systems integration with a practical understanding of complex business operations.
At DeepLight, we don't believe in "off-the-shelf" fixes. We deliver tailored AI solutions designed to integrate seamlessly into existing enterprise architectures, ensuring that innovation is both scalable and secure. From building robust data foundations to deploying sophisticated AI platforms, we empower our clients to lead in an increasingly automated world.
The Full Stack Developer builds the human-facing surfaces of the Agentic Workflow Automation (AWA) platform—specifically the Human-in-the-Loop (HITL) reviewer interface, the AgentOps operational dashboard, and internal AWA developer portals. Operating within the AI Centre of Excellence (AICoE) at a Band 7 / Mid-level capacity, this position works directly at the boundary between complex AI pipelines and end-users (including operations reviewers, business analysts, bank customers, and engineering teams). The role translates complex agent states, confidence scores, and exception data into high-performance, intuitive user interfaces that enable fast, informed human decisions.
Responsibilities
- HITL Reviewer Interface: Design and build the operations reviewer queue interface, incorporating high-resolution native document image displays, extracted field overlays with per-field confidence indicators, clear exception reasoning, and optimized keyboard-shortcut workflows.
- AgentOps Dashboard: Build and maintain real-time operational dashboards that consume Azure Monitor metrics and Log Analytics data to track STP (Straight-Through Processing) rates, HITL rates, token consumption, and latency metrics (P50/P95/P99).
- System & API Integration: Establish secure communication between AI core services and core banking apps via the AWA Conversation Orchestrator backend API. Connect applications to APIM-governed LLM endpoints and Orkes Conductor job status APIs.
- Internal Tooling & Portals: Build and maintain the AWA Use Case Manifest authoring tool, the Prompt Management System (PMS) UI, and ground-truth annotation interfaces used by the operations team for test data preparation.
- Performance & Accessibility Optimization: Ensure all web interfaces meet institutional accessibility standards, optimize application performance for lower-bandwidth branch locations, and implement graceful error handling for AI pipeline latency.
- Design System Maintenance: Build and maintain a unified, consistent AWA component library and design system across all interfaces, ensuring strict alignment with corporate brand guidelines.
