Tech Lead
RainsNowPresent · Remote · Full-time
typescriptvuenuxtgopostgresqldocker
// AI Engineer
Daffa Izzuddin
I take models from a training run to something people can actually use — train, serve, monitor, iterate.
- Status
- Open to AI engineering roles — remote, full-time
- Based in
- Malang, Indonesia · UTC+7
- Currently
Tech Lead · Rains
Founding Developer · Mayfeyr
- Model development
- Inference & serving
- Monitoring & iteration
- Shipping the product around it
I am an AI engineer who likes the unglamorous half of machine learning: getting a model out of the notebook, putting it behind a real endpoint, watching what it does to real users, and then going back to improve it.
Most of my work sits where ML meets web: image generation for batik patterns, computer vision on satellite imagery, face embeddings for authentication, and the Nuxt/Go/Postgres plumbing that makes those usable. I also write about the engineering side at invasikode.com, because explaining a system is a good test of whether I actually understand it.
What I actually do
Models are the easy part. These four stages are what I own end to end, and every claim above links to a project where I did it.
01Experience
Where I have shipped models and the systems around themToggle Experience
Founding Developer
MayfeyrNowPresent · Remote · Founding team
typescriptnuxtpostgresqlB.Sc. Computer Science
Universitas BrawijayaEducation · Malang, Indonesia
02Projects
Click a card for the technical detailToggle Projects
03Skills
Grouped by the ML lifecycle, each with evidenceToggle Skills
01
Data & Modeling
Framing the problem, cleaning the data, training and evaluating a model I can defend.
- Dataset preparation and label sanity checks
- Fine-tuning diffusion models (LoRA / SDXL) — GenBatik
- CNN / embedding models for vision — SPARSE, IdentifEye
- Evaluation beyond a single accuracy number — SPARSE
PyTorch Hugging Face Python NumPy
02
Serving & Inference
A model that only runs in a notebook is not a feature. I put it behind an endpoint and keep it honest.
- REST / async job APIs around model inference — GenBatik, IdentifEye
- Streaming progress to the client (WebSocket) — GenBatik
- Containerised deploy, GPU vs CPU trade-offs — SPARSE
- Latency and cost per request — GenBatik
FastAPI Go Docker Cloudflare
03
Monitoring & Iteration
The part most portfolios skip. I care about what happens after deploy: what broke, what drifted, and what the next version should change.
- Logging inputs/outputs to make failures reproducible — GenBatik
- Feedback capture (accepted vs rejected output) — TataKata
- Cost and usage tracking per feature
- Versioned experiments instead of one-off notebooks
Postgres Grafana GitHub Actions
04
Product Delivery
An AI feature is only as good as the interface around it. I build that part too.
- Nuxt / Vue front ends for data-heavy products — GenBatik, invasikode.com
- API design and schema work in Postgres
- Static-first builds, prerender, and deploy pipelines — Portfolio (this site)
- Technical writing that explains a system to non-authors — invasikode.com
Nuxt TypeScript Tailwind Postgres
04Writing
Selected posts · 100+ articles, ~1k readers weekly at invasikode.comToggle Writing
React, Vue, Angular, what to Decide?
InvasikodeMembandingkan Vue, Angular, dan React dari sisi trade-off nyata, bukan dari preferensi.
Read →What's Up Javascript? SoJS 2022
InvasikodeRangkuman hasil survei State of JS 2022 dari 40k responden dan apa artinya untuk pemilihan tooling.
Read →Guide: Using Node and Express with TypeScript
InvasikodeSetup Node + Express dengan TypeScript dari nol, termasuk konfigurasi build dan type checking.
Read →Make CMD-ing Fun Again with Starship
InvasikodeMemasang Starship prompt untuk PowerShell dan merapikan environment terminal sehari-hari.
Read →
// Contact