Aditya Firmansyah

Multi-tenant e-commerce platform built for independent merchants and modern dropshipping. Powers isolated merchant storefronts with host-based tenant resolution, dynamic OKLCH color token theming, sub-second product search, bulk CSV catalog operations, and automated SEO metadata generation.

• Architected a high-performance monorepo: Next.js App Router frontend paired with a FastAPI (Python 3.11) backend and PostgreSQL database with host-level tenant isolation behind Cloudflare proxying.
• Engineered merchant storefront customization with dynamic OKLCH color spaces, zero-network self-hosted Google font subsets for deterministic CI, and synchronized bilingual localization (ID/EN).
• Designed resilient data pipelines including atomic bulk CSV product imports with SQLAlchemy session protection, multi-provider payment reconciliation, and background worker queues.

Open-source civic trust and neighborhood treasury system built for Indonesian RT/RW communities. Provides public financial transparency with no login required, self-check dues portal for residents, multi-month arrears detection, batch/advance payment processing, and granular role-based governance.

• Architected multi-tenant backend in FastAPI (Python 3.11) with strict PIC-based dues ownership and role segregation (Admin/Bendahara vs. Ketua approval workflow), backed by 85+ automated Pytest tests and zero-dependency XLSX/CSV reporting.
• Engineered modern, accessible frontend using Vite, React 19, TypeScript, and Tailwind CSS — featuring civic-trust design language, responsive compact mobile metrics, and sub-second phone-number lookup portal for residents.
• Production-deployed to VPS with Nginx reverse proxy, systemd daemon isolation, Let's Encrypt SSL, and automated CI pipelines enforcing strict Ruff and TypeScript quality gates.

GainForge turns real workouts into RPG quests — users earn XP, level up a character, and grow four stats (STR, END, AGI, FLX) by completing daily strength, cardio, flexibility, and endurance sessions. Built solo end-to-end: product design, schema, API, frontend, and production deployment.

• Full Rust stack front-to-back — Leptos (WASM) frontend and an Axum backend share a single DTO crate, keeping types identical across client and server.
• Designed the quest/XP/leveling system and a 4-stat character model, plus an admin panel for managing training programs, exercises, and weekly schedules.
• Shipped and self-hosted on a VPS with Docker and PostgreSQL — 15 schema migrations, 100+ commits, with iterative mobile-UX polish (drawer navigation, responsive schedule grid).

Users build a profile (skills, bio, experience), type in a job title and description, and get a tailored, editable cover letter back. Originally built Upwork-specific with URL-based job scraping; broadened to a general-purpose tool once Upwork shipped its own in-house AI proposal writer, making an external Upwork-only tool redundant.

• User profile and job details feed a prompt builder that calls an LLM API to draft the letter, followed by a humanization pass that strips AI-writing tells (em dashes, buzzwords, promotional filler) before the user reviews and saves it.
• Full auth/profile/letter-history schema in PostgreSQL (users, profiles, projects, cover_letters with draft/saved status) behind JWT auth.
• AI backend (endpoint + model) is env-configurable rather than hardcoded — swapped providers in production without a code change when the original vendor's model was discontinued. Self-hosted via git-push deploy (bare repo + post-receive hook + systemd + nginx).