Production Projects

Production-grade AI, backend, data engineering, and full-stack systems built for real users with a strong focus on scalability, reliability, maintainability, and engineering quality.

Flipped.ai — Production AI Systems

RAG vLLM Document Intelligence Arabic OCR FastAPI GPU Inference

An engineering case study of the production AI behind an AI-native recruitment platform: document intelligence, multilingual/Arabic OCR, hybrid RAG retrieval, semantic candidate matching, and open-weight LLM serving with vLLM behind async, distributed microservices.

Engineering Highlights

• Document-intelligence pipeline for messy, multilingual CVs; custom Arabic OCR/extraction (~80% → 95%+ parsing accuracy)
• Hybrid retrieval (BM25 + dense) with reranking and MMR for semantic candidate matching
• Open-weight LLM serving with vLLM — continuous batching, PagedAttention, KV-cache management
• Async FastAPI + Celery/RabbitMQ/Lambda microservices with ATS integrations (Zoho Recruit, Greenhouse)

Read Case Study

People Case Management Automation Platform

FastAPI React/Next.js AWS RAG Agentic AI Memory LLM Evaluation Event-Driven Architecture

An AI-powered Legal Document Intelligence platform built for Centrica UK, combining agentic AI, advanced RAG, long-term memory, evaluation, and cloud-native event-driven infrastructure to automate complex HR and legal case-management workflows.

Engineering Highlights

• Led end-to-end delivery as a Forward-Deployed & Senior Innovation Engineer
• Architected a cloud-native, event-driven AI platform (FastAPI, Next.js/React, ECS Fargate, Lambda, SQS, S3, EventBridge, Docker, Terraform)
• Engineered advanced/agentic RAG and semantic long-term memory for large, growing case histories
• Built LLM evaluation and observability with DeepEval, RAGAS, regression testing, and Langfuse tracing
• Conducted AI security hardening and pre-penetration testing with the Strix security agent

Read Case Study

Centrica Spark — AI Innovation Platform

LLMs DSPy Context Engineering ReAct Autonomous Planning Multi-Agent Systems AWS AgentCore Evaluation

An internal Centrica innovation platform where LLMs and agentic workflows help employees develop ideas through structured research, contextual analysis, feasibility assessment, risk analysis, and actionable recommendations.

Engineering Highlights

• Architected agentic AI workflows using LLMs, DSPy, context engineering, ReAct-style reasoning, autonomous planning, and multi-agent orchestration
• Designed context-aware LLM pipelines combining employee ideas with enterprise knowledge, policies, and organizational constraints
• Engineered specialized agents for research, analysis, feasibility, risk, and recommendation behind an orchestration layer
• Designed long-running autonomous agent workflows using AWS AgentCore for stateful, multi-step execution
• Built end-to-end evaluation and observability with DeepEval, RAGAS, regression testing, and Langfuse tracing

Read Case Study

Project Blueprint — AI-Enabled Residential Digital Twin

LiDAR AutoCAD/CAD Point Clouds Three.js Digital Twin HVAC Agentic AI Context Engineering

An AI-enabled residential digital-twin system combining mechanical engineering, LiDAR scanning, CAD/AutoCAD data, 3D visualization, and agentic AI to digitally represent UK homes and their heating, cooling, and ventilation infrastructure for energy-aware optimization.

Engineering Highlights

• Architected a residential digital-twin workflow combining mechanical engineering, 3D reconstruction, and AI engineering
• Designed a LiDAR-to-digital-twin pipeline combining spatial scans with AutoCAD/CAD floor plans
• Designed interactive Three.js representations of residential spaces, HVAC systems, heat pumps, and ventilation
• Designed an agentic AI layer capable of reasoning over spatial, HVAC, and energy data
• Applied multimodal/context engineering to combine 3D geometry, CAD metadata, and HVAC configuration for grounded recommendations

Read Case Study

Dhokla House

React.js Node.js Express.js PostgreSQL MongoDB

Production food-ordering and brand platform focused on menu discovery, order intent capture, and conversion-optimized customer journeys.

Engineering Highlights

• Modular React.js UI components with responsive layouts
• Node.js + Express.js APIs with clean route separation
• Hybrid persistence: PostgreSQL for transactions, MongoDB for flexible content
• Optimized for rapid UI iteration and backend reliability

Visit Project

Apni Dukaan

React.js Node.js Express.js PostgreSQL MongoDB

E-commerce platform focused on catalog exploration, customer checkout, and operational order management in a production environment.

Engineering Highlights

• Reusable catalog and checkout components for friction-free purchase flows
• Clear separation between read-heavy catalog and transactional checkout routes
• PostgreSQL for order transactions, MongoDB for product metadata
• Maintainability and latency control through compact service boundaries

Visit Project