Mohammed Izhaar Ul Haq

NNE Engineering Model — AI-Powered Knowledge Graph PoC

• Built an end-to-end PoC replacing a flat Excel-based Engineering Model (1,800+ entries across 28 disciplines, 6-level hierarchy, 5 project phases) with an AI-powered, graph-backed system for navigating deliverables and cross-discipline dependencies.
• Designed a Neo4j knowledge graph modeling disciplines, phases, activities, and deliverables as typed nodes connected by relationships (CONTAINS, FEEDS INTO, EVOLVES TO, BELONGS TO), enabling multi-hop dependency traversal that flat spreadsheets or traditional RAG could not support.
• Built an AI conversational assistant translating natural language questions into graph queries, returning multi-hop answers and structured aggregations directly from the graph.
• Developed the full-stack visual layer end-to-end using Next.js and D3.js, delivering an interactive Hierarchy Navigator, a Cross-Discipline Dependency Graph, and a Phase Timeline.

TMM First Response Suite (AWS · Zendesk)

• Built the First Response Agent — an event-driven, serverless pipeline (API Gateway, SQS, Lambda, DynamoDB, Bedrock) that ingests Zendesk tickets and produces contextual first-response drafts via a 3-step LangChain prompt strategy with 33 category-aware templates. Reduced first-response SLA from ~4 hours to ~67 seconds while handling 60,000+ tickets per month, at $0.027–$0.069 per ticket.
• Built the Past Ticket Analysis RAG service — a fully serverless ingestion and retrieval pipeline using Amazon OpenSearch Serverless (vector engine), with two-tier retrieval (deterministic metadata match, then semantic fallback) and 17 category-specific lookup strategies.
• Designed the Merge Agent — a duplicate-ticket detection and merging workflow on AWS Step Functions, preserving PO/invoice numbers and cleaned error messages on the surviving ticket. Eliminated manual duplicate-merging work previously done by support engineers.
• Shipped two Zendesk sidebar apps (ZAF v2): a Feedback App with HMAC-signed AWS ingestion and a Chat App grounded in the RAG knowledge base with citation deep-links.
• Designed a daily SharePoint customer-handling-notes sync using EventBridge + Amazon AppFlow + S3, with an S3-triggered Lambda parser writing structured notes into DynamoDB for AI prompt enrichment.
• Built a Role-Based AI Operations Dashboard on AWS Fargate (ECS) surfacing AWS usage and AI response metrics (success rate, feedback split, latency, per-ticket cost), with role-based access control gating Business vs Developer views.
• Set up automated CI/CD pipelines on AWS CodeBuild and AWS CodePipeline integrated with GitHub across the suite, enabling push-to-deploy with sandbox-to-production promotion and automatic rollback on failure.

AI-Based Legacy Data Extraction & Processing Tool (MOSPI Feeder Platform)

• Developed full-stack document intelligence system for Ministry of Statistics, Government of India, serving 1000+ users with automated document analysis and processing.
• Built AI/ML pipeline with Docling, Vanna.AI, and Ollama for automated OCR, table extraction, and NL-to-SQL conversion — reducing document processing time by 80%.
• Integrated Qdrant vector database for semantic search and RAG-based chat; deployed on H200 GPU server with in-house hosted OpenAI OSS 120B model.

AI-Based Intelligent Search Solution (MOSPI Platform)

• Developed production-grade document intelligence and RAG chat system using Django and React, enabling natural language queries across multi-format files (PDF, DOCX, PPT, CSV).
• Built real-time chat with Django Channels WebSockets, streaming LLM responses from Ollama with voice transcription support.
• Implemented RAG pipeline from scratch with Qdrant vector database and Docling OCR for semantic search.

Hirenex Resume Screening

• Architected and developed complete resume-screening flow as a full-stack solution, integrating AI-powered resume analysis with NLP to automate candidate filtering based on job-specific criteria.
• Stack: Python, Django, Celery, PostgreSQL, Redis, Qdrant, React, Next.js, Docker, WebSockets, Docling, Vanna.AI, Ollama, OpenAI API.

AI Interviewing Tool

• Developed complete AI interviewing platform supporting one-way and two-way interviews with 20+ languages, customizable AI avatars, and comprehensive evaluation reports.
• Built adaptive interview logic where, after every answer, the AI checked coverage against a predefined set of interview criteria and dynamically generated targeted follow-up questions, ending the interview automatically once all criteria were covered.
• Built a post-interview scoring pipeline that passed the full transcript and criteria matrix to an LLM to generate a configurable score with per-question and per-criteria feedback.
• Integrated Azure AI real-time avatar system for interactive interview experiences with natural language processing.
• Built mock-interview platform and skill-assessment module with real-time feedback for candidate practice sessions.
• Developed a browser-based proctoring system from scratch using Google MediaPipe, running entirely client-side via WASM with no backend processing, to flag mobile phone usage, multiple people in frame, and multiple simultaneous voice sources.

EUCFY — EUC Risk Mitigation Platform

• Planned and designed complete workflows from scratch using Excalidraw, translating requirements into scalable frontend architectures.
• Developed EUC Risk Mitigation Platform using React to transition banks' Excel VBA Macros to secure web applications, implementing ABAC (Attribute-Based Access Control) from scratch.
• Built responsive UI components and landing pages using React.js, Next.js, and Tailwind CSS for cross-platform compatibility.