AI Engineer intern supporting Go-to-Market, translating enterprise discovery into Microsoft AI architectures, prototype demos, and business cases to deliver practical AI solutions.

• Partnered with Business Development on 4 enterprise AI engagements, translating discovery into Microsoft solution architectures, business cases, and live demos for procurement, document operations, analytics, and financial-services workflows
• Built a procurement and sourcing agent on Microsoft Foundry using SAP Cloud SDK to connect to the client's SAP ECC, S/4HANA, and Ariba APIs, with specialist review and an audit trail; validated 25 model calls on a curated 494-line procurement snapshot, producing 147 proposed RFQ groups in 147.6 seconds
• Developed a Python evaluation harness for multi-format invoice extraction, benchmarking Microsoft Foundry, Azure Content Understanding, and Azure Document Intelligence across 10 invoices and 3 repeated runs; recommended gpt-5.4 at 83% precision and recall, 8.9s median latency, and Rp792 per invoice
• Architected an automated IT-maintenance solution leveraging Azure SRE Agent to apply OS, dependency, and in-server software updates against compliance checklists, verifying each maintenance step before completion
• Architected a Microsoft Copilot solution in Microsoft Teams for division leads, connected to a Power BI model to deliver daily priorities and announcements, answer KPI questions, and provide conversational reporting without opening a dashboard

• Built the WhatsApp Customer Service AI Agent via the Meta WhatsApp Business API on the Cloudflare Agents SDK with 10+ tools, multimodal receipt reading, and a channel-agnostic schema; cut LLM cost by roughly 50% through prompt-cache optimization
• Built user-configured transaction recategorization rules using multi-stage similarity matching across substring, fuzzy, trigram, and Gemini Embedding 001; after 3 consistent corrections for the same merchant and category, Cloudflare Queues triggers an LLM suggestion of a reusable brand pattern for dialog approval
• Designed a merchant-enrichment workflow that identifies high-frequency transaction counterparties through a daily OpenClaw cron, then uses the Brave Search API and web scraping to create reviewable merchant context for downstream LLM categorization
• Extended the agent architecture to Telegram, enabling proactive budget alerts and conversational financial assistance alongside WhatsApp
• Built the Remix product landing page, made every user-facing workflow I developed available offline with PowerSync local-first sync, and instrumented telemetry and error monitoring with PostHog and Sentry dashboards

• Implemented zero-trust IAM across 3 microservices using Go, JWT RS256, and Redis, enforcing endpoint-level authorization and immediate session revocation
• Delivered the listing-management service with tenant defaults and overrides, product and variant CRUD, plus asynchronous CSV/XLSX bulk operations with job tracking
• Built a Shopee platform connection module with OAuth 2.0, encrypted token storage, caching, and background refresh, enabling multi-shop connection health and status visibility
• Established versioned database migrations and standardized API response contracts across services, improving schema-change safety and client integration consistency