Scroll to explore
Ibrahim Sharar
Projects

Things I've Built

Each project has a live interactive demo — click any card to try it out.

5 projects · All demos are interactive
View Project
reinforcement learning

DRL Blackjack AI Learns to Play Cards

I trained deep reinforcement learning agents (PPO and A2C) to play Blackjack and navigate multi step app flows autonomously. Each agent uses config driven YAML personas with different reward strategies, from aggressive to cautious to balanced. The demo below is a fully playable Blackjack game where you can hit, stand, and deal hands right here. The research side generated HTML reports with training plots and GIF replays.

  • PPO and A2C agents with YAML configured personas and reward weight tuning.
  • Reproducible training with fixed seeds, CSV episode logs, and JSON summary exports.
PythonStable Baselines3GymnasiumPyTorchYAML+1
View Project
graphics

GPU Global Illumination

A deep dive into physically based rendering. I built a real time shader pipeline that simulates how light actually bounces around a scene using Schlick Fresnel approximations and Monte Carlo sampling. The result is a cubemap environment where reflections, refractions, and diffuse light transport all play together in real time. Tweak the sliders and watch the physics change live.

  • Schlick Fresnel approximation for physically accurate reflection and refraction blending.
  • Irradiance maps and Monte Carlo integration (500 samples) for soft diffuse lighting.
C++OpenGLGLSLGLFWGLEW+1
View Project
product

MoveMate Movers Booking App

A full end to end moving company booking prototype I designed from scratch. The idea was simple: what if booking movers was as easy as ordering food? The app walks you through selecting your move details, comparing movers with transparent pricing, booking a date, tracking your move in real time, and leaving a rating after. Everything persists in localStorage so try booking a move right in the demo.

  • Six screen booking flow with inventory counters, dynamic pricing, and form validation.
  • Mover comparison cards with sorting, simulated GPS tracking, and post move star ratings.
HTML5CSS3Vanilla JSLocalStorage
View Project
LLM app

Micro Consultant AI Document Generator

Built for small businesses that can't afford consultants. This app uses a local LLM (Ollama) with retrieval augmented generation to draft professional SOPs, employee handbooks, and compliance policies. Upload your business context, pick a document type and tone, and it generates a tailored draft. I added input validation, prompt injection guards, and per request telemetry logging.

  • End to end FastAPI backend with RAG retrieval over uploaded business documents.
  • Safety layer with input length limits and prompt injection heuristic checks.
PythonFastAPIOllamaRAGVector Search+1
View Project
ops intelligence

OpsPilot Operations Intelligence Platform

An operations intelligence tool that ingests logs, metrics, and event streams to surface bottlenecks, anomalies, and hidden cost leaks. It uses a combination of rule based checks and lightweight ML heuristics to flag issues, then an LLM layer translates findings into plain English insights. Adjust the load slider in the demo to see how metrics respond under pressure.

  • Unified ingestion pipeline for logs, metrics, and live event feeds.
  • Heuristic + ML anomaly detection for latency spikes, failures, and cost drift.
Data PipelinesHeuristic AnalysisLLM SummariesDashboard UI