Ajinkya Dessai

<p>● Designed and prototyped LLM-powered AI workflows for consumer and enterprise use cases, using prompt engineering, tool-</p><p>routing rules, and evaluation rubrics to rapidly iterate and validate features aligned with business requirements and user stories.</p><p>● Developed interactive prototypes and visualised user journeys with Whimsical and personas via UXPressia, enabling data-informed</p><p>prioritisation of features and early testing with stakeholders.</p><p>● Applied AI experimentation frameworks with explicit success criteria, failure modes, and guardrails to assess the feasibility of</p><p>generative AI and AI agent automations, reducing consultation and feature validation cycles by 70% and informing compliant, risk-</p><p>aware rollout decisions.</p><p>● Translated early LLM and automation experiments into a roadmap for product adoption, aligning AI capabilities with business and</p><p>user needs, and prioritising agentic use cases based on customer pain points and competitive insights.</p><p>● Delivered insights from user testing and rapid experimentation that informed next-generation generative AI features, improving</p><p>early-adopter engagement and demonstrating scalable AI product thinking across onboarding and policy-like workflows.</p>

<p>● Promoted to lead core React, Typescript, and AWS stack, refactoring AWS architecture and APIs to centralise monitoring, tighten</p><p>security, and remove redundancies. Applied GenAI tool-routing to speed delivery while maintaining secure, compliant practices.</p><p>● Mapped user journeys, implemented API security and monitoring best practices, and coordinated cross-functional teams; reduced</p><p>ambiguity through LLM/GenAI-assisted drafts for user stories and documentation, using clear constraints, context management,</p><p>and acceptance criteria.</p><p>● Designed release flows and integrated automated testing for services and UI, raising test coverage to 92% and improving platform</p><p>reliability, using prompt-based test generation with LLMs and evaluation rubrics to expand coverage faster.</p><p>● Led cross-business implementation and delivery of a multimillion-dollar Microsoft-partnered IoT platform by defining end-to-end</p><p>ingestion paths and guiding adoption. Cut cloud spend by 30% and reduced compute costs by 20%, aligning execution with product</p><p>strategy.</p><p>● Built a hardware-aligned EV-charging simulator to remove integration bottlenecks, enabling one-touch deployments and</p><p>accelerating stakeholder validation, informing roadmap prioritisation.</p><p>● Designed release flows and integrated automated testing for services and UI, raising test coverage to 92% and improving platform</p><p>reliability for faster, higher-quality feature rollout.</p><p>● Redesigned metadata pipelines with durable Azure Functions and surfaced lineage in Purview, reducing redundant operations by</p><p>30% and improving downstream model traceability and auditability for data- and AI-powered features in a regulated, enterprise</p><p>context.</p><p>● Mentored junior engineers, setting best practices for cloud ML delivery, applied data science projects, and responsible use of</p><p>generative AI tools in production-like environments.</p>

<p>● Designed and validated a hybrid statistical forecasting system for optic failures, achieving 90%+ week-level accuracy and 6–8</p><p>months lead time. This proactive insight informed maintenance strategy and influenced product roadmap prioritisation.</p><p>● Developed ETL pipelines to improve analytics data availability, moving optics telemetry into Athena with SQL transformations.</p><p>Improved data accessibility and processing efficiency by 20%, enabling faster, data-driven product decisions.● Standardised ML deployments with one-command CI/CD flows and automated checks, reducing deployment errors to near zero</p><p>and cutting setup time by 15%. This ensured reliable feature delivery for cross-functional stakeholders.</p><p>● Prototyped an NLP-based anomaly detection scoring system using LSTM experiments, improving detection capability and providing</p><p>actionable insights for security feature planning.</p>