Software Engineer · Applied AI & Backend Systems
Python · APIs · Data · Machine Learning · Reliable Software
Email · LinkedIn · LeetCode · Google Cloud Skills Boost
I am a software engineer with experience across Python backend development, applied AI, databases, APIs, automation, and end-to-end product delivery. I enjoy understanding how a real workflow operates, turning its constraints into a clear technical design, and building a solution that remains understandable after the first version.
My projects range from enterprise knowledge systems and developer tools to industrial optimization and computer vision. Across domains, I care about the same fundamentals: correctness, maintainability, useful interfaces, thoughtful data design, and software that solves the intended problem.
| Project | What it demonstrates | Core technologies |
|---|---|---|
| API Migration Agent | Detects OpenAPI breaking changes, evaluates their impact on Python clients, proposes human-approved patches, and validates them in an isolated environment. | Python, FastAPI, LangGraph, LiteLLM, Gemini, Next.js |
| Omni AI Agent | Stateful WhatsApp and Discord agent with MCP-separated tools, persistent memory, semantic retrieval, and execution tracing. | Python, LangGraph, FastMCP, PostgreSQL, pgvector, Opik |
| Agentic AI Scaffold | Generates modular AI backends with explicit architectural boundaries and optional persistence, MCP, and observability integrations. | Python, FastAPI, LangGraph, SQLAlchemy, FastMCP, Opik |
| Ingestor Pro | Unified workspace for ingesting, exploring, and conversing with files, PDFs, spreadsheets, folders, and web content. | TypeScript, Python, Gemini, D3.js, document processing |
Earlier machine-learning work
- CV Helper — CV/job-description analysis and personalized feedback using NLP, retrieval, and AI-agent workflows.
- Face Sight — real-time face detection with a fine-tuned YOLO model and Streamlit interface.
- Text Flow — text correction and next-word prediction using n-grams and LSTM models.
Understand the real problem
↓
Make assumptions and constraints explicit
↓
Build the smallest coherent system
↓
Observe failures, evaluate behavior, and learn
↓
Strengthen the boundaries that make it reliable
- Useful before impressive — the architecture should serve a real workflow.
- Explicit before magical — tools, permissions, state, and failure paths should be understandable.
- Human judgment belongs in the design — especially when an automated action has consequences.
- Maintainability is a feature — a system should be easier to extend after delivery, not harder.
- Learning happens through delivery — unfamiliar technologies become useful when applied to a concrete problem.
| Area | Technologies |
|---|---|
| Software & backend | Python, FastAPI, Pydantic, SQLAlchemy, REST APIs, WebSockets, AsyncIO, OAuth2 |
| Data systems | PostgreSQL, pgvector, Supabase, relational modeling, SQL, document ingestion |
| Applied AI & ML | LangGraph, RAG, MCP, PyTorch, TensorFlow, Scikit-learn, OpenCV, YOLO |
| Delivery & observability | Docker, Git, AWS, GCP, Arize Phoenix, Opik |
| Interfaces | Next.js, TypeScript, JavaScript, Streamlit, D3.js |
- The ability to move from an unclear requirement to a working, demonstrable system.
- Strong Python foundations and practical experience connecting APIs, databases, AI components, and user interfaces.
- A learning mindset: I am comfortable entering an unfamiliar domain, identifying what matters, and becoming productive quickly.
- Ownership of the details that make software usable—data integrity, failure handling, security boundaries, documentation, and maintainability.
- Clear collaboration and a willingness to explain decisions, receive feedback, and improve the result.
I am open to engineering opportunities where I can learn, contribute, and help turn real requirements into dependable software—including international remote work and relocation.
Building systems where intelligence is useful because the engineering around it is dependable.