Hi, I'm aleph.
AI | Machine Learning | Backend | Cloud Infrastructure | Frontend | Mobile based in 🇩🇪.
Senior Full-Stack ML Engineer with 8+ years shipping end-to-end products. I specialize in owning technical projects from infrastructure architecture to polished user interfaces, with deep expertise in ML systems and AI integration.
I write about tech, language learning, and life on my blog.
Work highlights
AI & ML
-
Recommendation & personalization engine - Multi-territory pipeline across 12+ markets, evolved from Spark/TensorFlow to GPT-2 next-deal prediction, BERT4Rec, and NCF embeddings with MMR diversification
-
Agentic AI travel assistant - Multi-agent LLM orchestration with real-time streaming, shipped to production
-
RAG assistant & eval platform - Marketplace conversational AI with pgvector retrieval, MMR reranking, streaming tool calling, LLM-as-judge evaluation framework, and operational quality dashboards
-
Sentiment analysis pipeline - Customer feedback processing at scale, reduced manual review by 70%
-
NLP & content feature engineering - Automated NLP content features, GIS enrichment, Item2Vec embeddings, and content similarity scoring with t-SNE monitoring dashboards
-
Computer vision system - Real-time dice and player move detection from video streams
Infrastructure & Tooling
-
Real-time ML serving - Prediction APIs handling 1M+ requests/day with low-latency inference
-
GPU autoscaling platform - Kubernetes workload scaling with Argo Workflows, cut compute costs ~50%
-
Cloud resource CLI tools - Custom CLIs to scale, manage, and cost-optimize GPU training, inference, and ETL workloads
-
Production cloud infrastructure - ECS Fargate, RDS, ALB with Terraform IaC across multiple environments; 99.9% uptime
Featured Projects
kotoba 言葉
Full-stack AI language learning platform. Architected the backend APIs in FastAPI, deployed ML inference pipelines, built the React Native iOS app, and Next.js marketing site. Download on iOS
- FastAPI
- React Native
- Next.js
- Firebase
- LLM Inference
nagare 流れ
Hyper-lean, single-binary DAG orchestrator and workflow engine. A lightweight alternative to Airflow with zero external dependencies, YAML-defined pipelines, and a built-in web UI.
- Go
- TypeScript
- Next.js
- SQLite
- Docker
ratchet
A tool for giving AI coding agents a very particular kind of job: improve a measurable result, one iteration at a time. You hand that file to any AI coding agent and it runs an autonomous loop.
- Rust
- Coding Agents
- AI
fastapi+docker boilerplate
Cookiecutter template for FastAPI with Docker. Includes dev/prod/test configurations, Makefile shortcuts, environment-based settings, and uvicorn auto-reload for streamlined development and deployment.
- FastAPI
- Docker
- Docker Compose
- Cookiecutter
- Uvicorn
Also worth a look
-
dependia - Python dependency management tool with visual tracking dashboard
-
my japanese level - Japanese text analysis tool with custom kanji segmentation
-
bardo ai summarizer - Rust API for content extraction and GPT-3 summarization
-
python way - Curated collection of Python learning resources with 100+ links
Languages
I love learning languages and will know the most random things in the most random languages at different levels.