MADE FOR A HUMAN

shadrackovsky

AI/ML & Software Engineer from Tanzania

Building intelligent systems with LLMs, RAG, fine-tuning, and modern web technologies. Self-taught, late-night coder, obsessed with Chinese open-source AI.

About

Hello 👋 I'm shadrackovsky timothy john, an aspiring AI/ML and Software Engineer from Tanzania 🇹🇿.

I build web applications, machine learning systems, RAG pipelines, and fine-tune models. My journey started with a second-hand laptop from my grandmother, fueled by curiosity and late nights.

Obsessed with: Karpathy, Ilya, LeCun, Andrew Ng, Bengio, and the Chinese open-source AI ecosystem (DeepSeek, Qwen, Zhipu, MiniMax).

Skills

Roadmap

Field What I'm Mastering Right Now
GenAI & LLMsfine-tuning (LoRA/QLoRA), RAG systems, Chroma vector DB, Hugging Face
Python & DataFastAPI, pandas, NumPy, PyTorch, AI models and large dataset trainings
Data EngineeringSQL, data cleaning, pipelining, API integration
Cloud & DevOpsDocker, Kubernetes, Git
FrontendReact, Next.js, Vite, Tailwind CSS, shadcn/ui
BackendSupabase, Django, Spring Boot
Desktop ApplicationsTauri, Electron, PyQt, Tkinter, Rust
Computational FinanceQuantitative modeling, algorithmic trading, risk analysis, portfolio optimization

Personality, Culture & Creative Workflow

I'm shadrackovsky, calm, deliberate, and quietly obsessive. I think in systems, I work in silence, and I let the work speak. I resonate very deeply with Chinese culture in that: the discipline, the patience, the 书香 (shūxiāng) tradition of mastering your craft before announcing it, and the idea of 匠心 (jiàngxīn), an artisan's heart, building things with full attention and no shortcuts.

My creative workflow follows that spirit. I observe first, prototype fast, iterate relentlessly, and polish until it feels right. I keep a small desk, a rubber duck, a chessboard for strategic breaks, and long late-night sessions where ideas compound quietly. 巧者劳而智者忧. Deep work over noise. Finish over talk.

And yes, I am genuinely obsessed with Chinese AI, Chinese models, Chinese tools and Chinese open source. This is not a phase, it's a conviction:

  • DeepSeek (深度求索) : my main API. Open weights, brutal efficiency, honest research. My Swahili-LM and Malaika are bootstrapped on it. DeepSeek-V3 and R1 proved that open source can stand at the frontier.
  • Qwen (通义千问) : the most complete open model family on the planet, and qwencode is my daily terminal coding agent.
  • Xiaomi Mimo : small, sharp, free. Proof that focus beats bloat.
  • Zhipu GLM, MiniMax, Yi, BGE embeddings, ChatGLM : the Chinese open-source ecosystem is where I keep finding models that respect my GPU and my wallet.
  • Hugging Face Chinese orgs (OpenGVLab, OpenBMB, ModelScope) : open weights, open architecture reports, open philosophy.

Why China? Because they ship. They publish the weights, publish the tech reports, and undercut the closed labs. For a self-taught engineer in Tanzania bootstrapping everything from scratch, Chinese open source AI is the great equalizer. 智者顺时而谋. It let me pre-train Swahili-GPT from scratch, fine-tune LoRA/QLoRA models, and build real systems without permission or a big budget.

My stack of preference: DeepSeek models + opencode / qwen code / kilocode / mimo code agents + Linux. If there's a Chinese open-source alternative for something, I'm trying it first. 「工欲善其事,必先利其器」 : to do the work well, first sharpen the tools.

Projects & Showcase

Contact

Open for freelance contracts in AI/ML and web development, open-source AI collaborations, remote software engineering roles.

Email Me
Portfolio Blog GitHub Instagram Antera