AI engineer and senior full-stack developer building production-style systems across responsible AI orchestration, agentic AI, RAG, MCP, MongoDB intelligence, data-driven workflows, contact-centre automation, and enterprise application architecture.

My work focuses on practical AI engineering rather than hype. I build systems that are structured, testable, grounded, and easy to review, with an emphasis on real architecture patterns such as orchestration, document ingestion, retrieval pipelines, provider abstraction, evaluation, traceable outputs, and full-stack product delivery.

This GitHub portfolio brings together applied AI engineering projects, full-stack architecture demos, review intelligence tools, RAG systems, MCP foundations, and practical automation utilities.

This portfolio shows how modern AI and software systems should actually be built.

The main themes across my repositories are:

• responsible AI orchestration, guardrails, evaluations, and release governance
• contact-centre and customer-service AI workflow simulation
• agentic orchestration and workflow design
• retrieval-augmented generation with grounded answers
• MCP and tool-ready architectures
• MongoDB-connected AI systems
• document and data ingestion pipelines
• sentiment, review, and customer intelligence workflows
• evaluation, traceability, and practical utility
• enterprise-grade .NET and Angular architecture
• Streamlit, FastAPI, and Dash apps that are easy to run and inspect

This is not a collection of toy chatbots. It is a portfolio of applied AI engineering, product architecture, and practical business intelligence work.

Responsible Contact-Centre AI Lab for safe customer-service AI orchestration in a regulated utility-style environment.

This project demonstrates how a production-style customer-service AI system can be structured, tested, observed, and released safely. It combines FastAPI orchestration, versioned assistant instructions, local deterministic model behaviour, optional AWS Bedrock support, local RAG, mocked business tools, input and output guardrails, Amazon Lex-style Lambda fulfilment, Amazon Connect-style handoff routing, golden-case evaluations, CI validation, Docker quick start, prompt-version comparison, and journey-level observability reports.

The project is designed as a fictional public-safe demo. It does not use real customer data, real supplier policy, real payment systems, or production credentials.

Core concepts:

• responsible AI orchestration
• regulated customer-service workflows
• guardrails for vulnerability, complaints, emergencies, financial actions, PII, and prompt injection
• retrieval from approved local knowledge articles
• mocked CRM, billing, meter, appointment, payment, and complaint tools
• Amazon Lex-style and Connect-style contact-centre patterns
• optional AWS Bedrock provider abstraction
• golden-case evaluations and regression tests
• prompt-version comparison for release governance
• local observability and journey metrics reporting
• GitHub Actions CI validation
• Docker and Docker Compose quick start

Flagship production-style boilerplate for grounded, voice-enabled agentic AI applications.

Built with Streamlit, FastAPI, MongoDB, Redis, versioned APIs, document ingestion, retrieval flows, voice input and output, and multi-provider support including mock, OpenAI, Anthropic, and Gemini patterns. Designed to demonstrate the shape of a serious AI platform rather than a single-purpose demo.

Core concepts:

• agentic application architecture
• RAG and grounding
• voice workflows
• document upload and chunking
• provider abstraction
• Redis and MongoDB integration
• API versioning and traces
• multi-provider LLM patterns

Homeprotect Review Intelligence Dash app for analysing Trustpilot-style review data.

This project processes customer review CSV data and turns it into structured insight across business areas such as Claims, Pricing, and Customer Servicing. It focuses on segmentation, sentiment, inferred NPS-style insight, key themes, and dashboard-ready outputs for stakeholder review.

Core concepts:

• customer review intelligence
• CSV ingestion
• segmentation logic
• sentiment analysis
• NPS-style inference
• business theme extraction
• Plotly Dash dashboarding
• stakeholder-ready analytics

Multi-document PDF RAG system with evaluation and page-aware citations.

This project converts PDFs to Markdown without relying on an LLM, chunks content for retrieval, stores embeddings, supports grounded multi-document chat, and evaluates retrieval quality with metrics such as MRR, Hit at k, Recall at k, Precision at k, and MAP at k.

Live demo: https://pdfragchat.streamlit.app/

Core concepts:

• PDF ingestion
• Markdown conversion
• chunking and embeddings
• page-aware citations
• multi-document RAG
• retrieval evaluation
• grounded answers

MongoDB export analysis and grounded insight generation.

MongoLens connects to MongoDB data, exports collections, generates profiling and analysis, and enables grounded Q and A backed by exported content. It is designed to bridge raw database exports and explainable insight workflows.

Live demo: https://mongolens.streamlit.app/

Core concepts:

• MongoDB inspection
• export and profiling workflows
• grounded answers over data
• RAG over structured exports
• data exploration
• citation-aware responses

MongoDB-focused MCP boilerplate exploring how model context protocols and database-connected tooling can work together in a structured AI system.

This repo is aimed at building clean foundations for AI workflows that need direct access to MongoDB-backed resources, tool execution, and future-ready agent integrations.

Core concepts:

• MCP-oriented architecture
• MongoDB connectivity
• tool and resource patterns
• AI boilerplate foundations
• tool-ready assistant workflows
• database-backed AI systems

Agentic system template for orchestrator and worker services.

This project explores scalable agentic patterns rather than one-off prompting. It is designed for extensible automation workflows, task routing, tool use, service decomposition, and systems that can be containerised and evolved into larger AI platforms.

Core concepts:

• orchestrator and worker design
• modular agent workflows
• scalable service patterns
• automation-ready architecture
• semantic RAG patterns
• distributed agentic design

Review intelligence for hospitality and experience-focused businesses.

GuestPulses processes CSV and Excel review datasets, maps columns, extracts sentiment and emotional signals, classifies hospitality categories, and produces structured insight outputs for business use.

Live demo: https://guestpulses.streamlit.app/

Core concepts:

• sentiment and emotion analysis
• hospitality review intelligence
• CSV and Excel ingestion
• column mapping
• category classification
• exportable business insights

Enterprise-grade, full-stack event management platform demonstrating Clean Architecture, Domain-Driven Design, SOLID principles, and scalable product architecture.

The project uses a .NET Web API backend and Angular frontend to show production-style thinking across backend, frontend, API design, domain modelling, and maintainable enterprise application structure.

Core concepts:

• Clean Architecture
• Domain-Driven Design
• .NET Web API
• Angular frontend architecture
• enterprise application structure
• scalable backend design
• full-stack product engineering

Lightweight trading AI assistant demo designed for fast product review and stakeholder walkthroughs.

The project demonstrates how a simple AI assistant interface can support trading-style summaries, Q and A flows, weekly briefings, prompt-driven analysis, and optional OpenAI-backed responses.

Core concepts:

• AI assistant demo
• Streamlit interface
• trading-style summaries
• weekly briefing workflows
• question and answer prompts
• lightweight product prototyping

Utility for downloading and organising images from URL lists.

A practical supporting tool for dataset building, content workflows, and automation scenarios where batch image collection is needed quickly and reliably.

Core concepts:

• image downloading workflows
• dataset support utilities
• content pipeline tooling
• automation helper scripts

Public GitHub profile and AI portfolio landing repository.

This repository acts as the main portfolio entry point, bringing together applied AI engineering projects across agentic orchestration, RAG, MCP, MongoDB intelligence, document pipelines, evaluation, traceability, Streamlit, FastAPI, Dash, and full-stack product architecture.

Core concepts:

• AI portfolio structure
• public project positioning
• technical profile presentation
• recruiter-facing engineering summary
• GitHub profile documentation

Public fork of Microsoft's Generative AI for Beginners learning repository.

This is used as a study and reference fork for ongoing generative AI learning. It is included honestly as a forked educational resource, not as an original project.

Core concepts:

• generative AI study
• learning reference
• prompt engineering foundations
• educational material
• ongoing AI development practice

I build AI and software systems that deal with the hard parts properly.

That includes:

• agentic workflows with orchestrators, workers, tools, and gateways
• RAG systems with real ingestion, chunking, embeddings, grounding, and citations
• document pipelines for PDFs, database exports, CSV, and Excel
• MongoDB-connected intelligence workflows
• MCP-ready foundations for tool-using AI applications
• review intelligence and sentiment analysis systems
• dashboard applications for stakeholder insight
• enterprise-grade .NET and Angular applications
• evaluation and observability patterns to make results measurable and reviewable
• demos and boilerplates that are useful for learning, extending, and technical review

My public work regularly uses:

• Python
• Streamlit
• FastAPI
• Plotly Dash
• MongoDB
• Redis
• LangChain
• OpenAI APIs
• Anthropic and Gemini provider patterns
• vector stores and retrieval pipelines
• Docker and Docker Compose
• .NET Web API
• Angular
• structured API contracts and service layers
• AWS Bedrock patterns
• Amazon Lex and Amazon Connect-style workflows
• GitHub Actions
• Docker and Docker Compose
• evaluation and observability reporting

• PDF RAG Chat: https://pdfragchat.streamlit.app/
• GuestPulses: https://guestpulses.streamlit.app/
• MongoLens: https://mongolens.streamlit.app/

The strongest areas represented in this GitHub portfolio are:

Projects such as AI Superjack Foundry, PDF RAG Chat, MongoLens, Agentic OS, and AI Superjack MCP Mongo explore how AI systems can be grounded, structured, tested, and connected to real tools and data sources.

Several projects focus on retrieval, document ingestion, data exports, citations, and evaluation. The goal is not just to generate answers, but to generate answers that can be traced back to source material.

Projects such as Homeprotect and GuestPulses focus on turning unstructured review data into useful business insight, including segmentation, sentiment, key themes, customer pain points, and dashboard-ready outputs.

ApexEventFlow demonstrates senior full-stack engineering through .NET, Angular, Clean Architecture, and Domain-Driven Design. This complements the AI portfolio by showing strong software engineering foundations beyond AI prototypes.

Projects such as Grabapic and the MCP Mongo boilerplate show practical automation thinking, tool execution patterns, and reusable foundations for larger systems.

python -m venv .venv