MUTTAQINMUZAKKIR

I learned to pair fast AI-assisted shipping with product judgment, code review, and explicit usage controls.

• Built tenant-aware AI quota and billing enforcement using per-school Durable Objects, feature entitlements, and provider-side preflight checks.
• Added reservation and idempotency paths so over-budget or unentitled workloads fail explicitly before provider calls.
• Built seeded school scenarios, impersonation, and E2E fixtures so demos and QA could use realistic teacher and admin workflows.

I owned the technical backbone of a company-facing game-based assessment product with a very small engineering team.

• Built a staging-first blue/green deployment system on GCP with immutable image promotion, automated smoke tests, rollback, and environment drift checks.
• Built a custom Sentry bridge from an HTML5 GameMaker runtime into browser-side observability with live session context and connection-failure telemetry.
• Built an async export engine that normalized gameplay events and survey responses into SPSS-ready reports for validity-study preparation.

Aiccountant raised my bar for craftsmanship: fast agentic coding only mattered when paired with reliable ownership.

• Implemented client-side encryption for locally stored data and aligned it with the existing envelope-encryption architecture to unblock compliance work.
• Replaced Gmail polling with a Pub/Sub push pipeline so onboarding users reached live inbox sync much faster.
• Shipped customer-facing referral, bulk transaction editing, and shareable insights features alongside the platform infrastructure work.

Hiff taught me how to make ambiguous AI POCs concrete enough for stakeholders to judge.

• Built a multi-stage pipeline using PDF parsing, Gemini Vision, and structured JSON extraction for a dense 400+ trademark document.
• Delivered a trademark similarity analysis tool across FastAPI and React for client-facing demos.
• Turned time-heavy manual review into a repeatable AI-assisted workflow that could be validated with stakeholders.

I turned an ambiguous client conversation into a concrete booking product, then owned the implementation end to end.

• Built fail-open Google Calendar sync for bookings on Cloudflare Workers using a Postgres outbox, queue-backed delivery, encrypted OAuth storage, and verified webhooks.
• Implemented slot locking and approval-based rescheduling to reduce double-booking across client, psychologist, and admin flows.
• Drove iterative product clarification with the client so UI, operations, and backend behavior stayed aligned during delivery.

This early project taught me to own the unglamorous parts of making messy stakeholder data usable.

• Standardized scattered survey records from roughly 50 companies across 6 sectors with normalization and fuzzy matching.
• Built a Django and Python visualization pipeline on a GCP VM that reduced reporting workflow from days to minutes.
• Containerized ingestion and analysis workflows with Docker to reduce environment mismatch and deployment friction.

KOMPeK was an early end-to-end ownership project where admin tooling mattered as much as the public flow.

• Built committee-facing workflows for registration monitoring, payment verification, analytics dashboards, and CSV exports.
• Designed secure submission and document flows with S3 presigned uploads, JWT-gated competition submissions, and protected asset retrieval.
• Delivered a system that served both public registration traffic and internal committee operations.