Projects & Portfolio
A focused selection of open-source infrastructure and AI tooling. Each project demonstrates an engineering decision, not just a technology list. Explore all code on GitHub.
🚀 Selected Projects
NTC Credit Risk Analyzer
Problem: make credit-risk signals more transparent for new-to-credit underwriting and behavioural auditing.
Approach: an explainable AI diagnostic framework with SHAP-based insights, event-driven design, and an interactive Streamlit experience.
Demonstrates: applied AI, explainability, and domain-aware product thinking.
Approach: an explainable AI diagnostic framework with SHAP-based insights, event-driven design, and an interactive Streamlit experience.
Demonstrates: applied AI, explainability, and domain-aware product thinking.
#Python #CreditRisk #ExplainableAI #SHAP #Streamlit #AWS #EventDriven
AI Query Router
Problem: route AI tasks to the least expensive model that can still produce a capable answer.
Approach: an autonomous routing agent that selects models at runtime, with Docker, Ollama, and GGUF-based local model support.
Demonstrates: model selection, token efficiency, and AI cost optimisation.
Approach: an autonomous routing agent that selects models at runtime, with Docker, Ollama, and GGUF-based local model support.
Demonstrates: model selection, token efficiency, and AI cost optimisation.
#AIEngineering #ModelRouting #TokenEfficiency #FireworksAI #Ollama #Docker
3-tier Todo API on EKS
Problem: deploy a repeatable, production-shaped API platform.
Approach: FastAPI on EKS, ALB ingress, and Terraform-managed infrastructure.
Demonstrates: Kubernetes delivery, routing, and modular IaC.
Approach: FastAPI on EKS, ALB ingress, and Terraform-managed infrastructure.
Demonstrates: Kubernetes delivery, routing, and modular IaC.
#AWS #Kubernetes #Terraform #IaC #EKS #FastAPI #ELB #k8s-elb-controller #ModularInfrastructure
Stateful AI Chatbot with Amazon Bedrock
Problem: preserve useful AI context without uncontrolled token cost.
Approach: FastAPI, Bedrock, MongoDB/DynamoDB persistence, sliding-window trimming, and per-call cost tracking.
Demonstrates: stateful AI architecture and operational cost control.
Approach: FastAPI, Bedrock, MongoDB/DynamoDB persistence, sliding-window trimming, and per-call cost tracking.
Demonstrates: stateful AI architecture and operational cost control.
#AWS #Bedrock #FastAPI #MongoDB #DynamoDB #Chatbot #Python #StatefulAI #CostTracking
Bedrock Model Compare
Problem: choose foundation models using comparable cost and latency evidence.
Approach: a CLI that runs one prompt against selected Bedrock models and writes a cheapest-first Markdown report.
Demonstrates: practical FinOps for generative AI.
Approach: a CLI that runs one prompt against selected Bedrock models and writes a cheapest-first Markdown report.
Demonstrates: practical FinOps for generative AI.
#AWS #model-benchmarking #cost-optimization #ai-tool #Python #boto3 #Utilities #SDK #Bedrock #library
More Open-source Work
Bedrock Cost Tracker Library
A non-invasive Python library for automated Amazon Bedrock cost tracking. Uses a decorator pattern to log token usage and USD cost to a local SQLite database for every call. Integrates with `bedrock-tui-helpers` for real-time pricing lookups and supports per-project cost segregation, making it ideal for monorepos or multi-service environments.
#AWS #Bedrock #Python #CostTracking #SQLite #Automation #CloudCost
Bedrock TUI Helpers
Shared terminal UI components and Bedrock model data for Python CLI tools — one pricing table, one model catalogue cache, one interactive params wizard, one model picker. This package is the single source of truth for all Bedrock-related terminal tooling in this repo. Both bedrock-model-compare and the chatbot cost tracking depend on it. Add it to any project to get a full model catalogue, pricing, and interactive prompts for free.
#AWS #ai-tool #Python #boto3 #Utilities #SDK #Bedrock #library
Boto3 Helpers
Reusable AWS SDK utilities for any Python project — eliminates repetitive boto3.client() calls, centralises region and credential resolution, caches clients at process scope, and provides idiomatic thin wrappers for Bedrock's control-plane and runtime APIs. Install once as an editable sibling package and every project in the repo gets consistent boto3 behaviour, a single mock-injection point for tests, and free connection reuse on warm Lambda invocations.
#AWS #Python #boto3 #Utilities #SDK #Bedrock #library
Multi-stack CloudFormation + Terraform IaC Module
A structured AWS infrastructure-as-code reference that demonstrates a hybrid provisioning pattern: Terraform as the orchestration layer driving CloudFormation stacks for core AWS resources. It encapsulates reusable templates for VPC setups, subnet segmentation, compute and storage services, and IAM roles with clean separation of concerns and modular design. This pattern supports production-aligned architecture and clear resource ownership across teams, enabling scalable, maintainable cloud deployments.
#AWS #CloudFormation #Terraform #IaC #VPC #EC2 #S3 #ModularInfrastructure
CI-driven Portfolio Platform
A fully automated portfolio platform built with static HTML, Python automation scripts, and GitHub Actions. This repository powers my personal site, generating pages and content from YAML configurations and external sources (badges, certifications, blog posts) and deploying via GitHub Pages. Showcases automation, scripting, CI/CD, and platform engineering in a real end-to-end system.
#HTML #CSS #JavaScript #Python #GitHubActions #GitHubPages