SELECTED WORK // 2019–NOW[DOC: MMB-WRK-2025 // 12-COL MONOGRAPH]

SYS.SPEC: VERIFIED ARCHIVAL RECORDS·SYSTEM VERIFIED

Work That Moves From Interface to Intelligence.

A verified collection of AI products, full-stack applications, enterprise SaaS platforms, machine-learning projects, automation workflows, and product-design case studies.

AUDIT MANDATEVERIFIED 1.0

Each project identifies my role, what was actually completed, and the current maturity of the work. No theoretical mockups represented as live software.

PORTFOLIO_LATENCY: OPTIMIZEDCOORDINATES: 33.6844° N, 73.0479° E

DISPLAYING 12 VERIFIED SPECIMENS

SPECIMEN ID: 01 // AQL-ENTERPRISE-AI-WORKSPACE/YEAR: 2026/ROLE: AI PRODUCT DESIGNER AND PROTOTYPE DEVELOPER

[ACTIVE PRODUCT]

CATEGORY: AI Product

AQL Enterprise AI Workspace

AQL explores an enterprise AI workspace connecting Google Drive, Notion and Jira. Its product architecture separates a Knowledge Agent for permission-aware retrieval, a Task Agent for Jira operations and a Workflow Agent for bounded cross-tool actions. The public repository currently shows an early frontend foundation, so connector execution, permission filtering and BGE reranking must be described as target architecture rather than completed production functionality.

Next.jsReactTypeScriptEnterprise UXRAG ArchitectureBGE RerankerGoogle DriveNotionJira

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 02 // DOCKET/YEAR: 2026/ROLE: FOUNDER AND PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Docket by ThinkSplice

Docket is a GPT-native project and task-management product. Users can create and update work conversationally while retaining a visual board and list for inspection and manual control. The plugin and dashboard share one backend and database, so the assistant is an interface to the same records rather than a separate source of truth. The alpha supports authenticated MCP tools, persistent user-owned projects and tasks, and optimistic version checks that reject stale updates.

MCPCloudflare D1Drizzle ORMTypeScriptVinextManaged AuthenticationJSON-RPCKanban UX

Read Case Study Live System

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 03 // MIMAR-STUDIOS-PLATFORM/YEAR: 2026/ROLE: FRONTEND AND PRODUCT DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

mimAR Studios Digital Platform

A responsive Next.js implementation translating approved desktop and mobile visual references into one semantic website. The build uses structured content, strict design tokens, an adaptive portfolio grid and a custom WebGL spatial effect. Performance safeguards include device-pixel-ratio clamping, IntersectionObserver pausing, tab-visibility handling and reduced-motion support. The implementation was checked across mobile, tablet and desktop breakpoints rather than maintained as separate websites.

Next.jsReactTypeScriptTailwind CSSWebGLResponsive DesignAccessibilityPerformance Engineering

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Localized visual assets

Reference viewports

1440 px and 390 px

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 04 // TIDYOS/YEAR: 2026/ROLE: AI PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

TidyOS

TidyOS is a local-first desktop agent for understanding, retrieving and organizing files. It extracts text from documents and images, combines semantic search with SQLite FTS5 and proposes useful names and destinations. Its central design problem is safety: it detects protected repositories, restricts operations to approved roots, records file actions in an audit ledger and supports rollback. Background monitoring uses stabilization checks so partially downloaded files are not processed prematurely.

PythonSQLiteSQLite FTS5LangGraphWatchdogDocument ExtractionOCRVector SearchDesktop Automation

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Automated tests

Implemented phases

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 05 // TRIPHISAAB/YEAR: 2026/ROLE: ANDROID AND AI PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

TripHisaab

TripHisaab explores expense tracking in the conversational interface people already use while travelling. The Android application pairs a WhatsApp interaction layer with a local Room ledger. Groq-assisted parsing converts natural-language messages into structured expenses, while deterministic accounting rules retain control of amounts, balances, dated entries, corrections and duplicate handling. Location-aware summaries help users understand spending across different stops without making cloud storage the default source of truth.

KotlinJetpack ComposeAndroidRoomGroqNotification ListenerLocal-first ArchitectureStructured Expense Parsing

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 06 // VERDICT-AI/YEAR: 2026/ROLE: AI PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Verdict AI

Verdict AI was built to help product teams move from scattered qualitative feedback to structured, actionable insight. The prototype focuses on extracting signal, grouping related feedback and presenting direct recommendations. Its current value is the product framing and working analysis flow; it does not claim validated research quality, team collaboration, billing or production-scale ingestion.

Next.jsReactTypeScriptSupabaseGroqStructured LLM OutputProduct Analytics UX

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 07 // CARDIOSCAN-AI/YEAR: 2026/ROLE: MACHINE LEARNING ENGINEER AND FULL-STACK DEVELOPER

[ACTIVE PRODUCT]

CATEGORY: AI Product

CardioScan AI

CardioScan AI explores computer-assisted analysis of paired stress and rest MPI SPECT images. The research compared VGG16, ResNet50V2 and DenseNet121 transfer-learning backbones on a six-channel representation formed from the paired images. The web system combines a Next.js clinical dashboard, a FastAPI and TensorFlow inference service, vessel-aware LAD, LCX and RCA outputs, Supabase history and a Groq-generated clinical-style report with a deterministic fallback. It is explicitly presented as research software, not a medical device or diagnosis system.

Next.jsReactTypeScriptTailwind CSSPythonFastAPITensorFlowKerasVGG16ResNet50V2DenseNet121SupabaseGroqHugging Face SpacesDocker

Read Case Study Live System Repository

EMPIRICAL METRICS & EVIDENCE

Original data

724 stress and 724 rest images

Augmented data

Approximately 2,000 paired samples

Model comparison

3 CNN backbones

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 08 // MAHAD-AI-PORTFOLIO/YEAR: 2026/ROLE: AI PRODUCT ENGINEER AND SYSTEM DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Mahad AI Product Engineering Portfolio

Instead of attaching a generic chatbot to a static portfolio, Mahad designed the portfolio as an AI product. Sanity owns approved content; deterministic ingestion writes canonical documents and chunks to Neon PostgreSQL and a rebuildable vector index to Qdrant. FastAPI serves retrieval and a LangGraph assistant, while an in-process ONNX classifier routes requests before expensive generation. The text assistant streams cited answers through SSE. A persistent ElevenLabs conversational companion is the next channel, sharing the same grounded knowledge and capped to two-minute voice sessions.

Next.jsReactTypeScriptTailwind CSSFastAPIPythonSanityNeon PostgreSQLQdrantmultilingual-e5-smallLangGraphONNX RuntimeMLflowLangSmithGroqSSEDockerElevenLabs

Read Case Study Live System Repository

EMPIRICAL METRICS & EVIDENCE

Router test accuracy

85%

Router p95 latency

16.77 ms

Assistant refusal score

100%

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 09 // BUSYFILE/YEAR: 2024–2026/ROLE: FOUNDING PRODUCT DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Busyfile

Busyfile began as a direct-to-consumer business-formation product and evolved into a B2B white-label platform that accounting firms could offer under their own brands. Mahad led product design across client onboarding, firm administration, internal operations and super-admin experiences. The system grew beyond formation into bookkeeping, tax and compliance services, requiring reusable patterns, role-specific information architecture and a design system that could scale without turning each service into a separate product.

FigmaProduct StrategyProduct DiscoveryUser ResearchInformation ArchitectureDesign SystemsInteractive PrototypingFintechEnterprise SaaSMulti-tenant Product Design

Read Case Study

EMPIRICAL METRICS & EVIDENCE

Interface scope

500+ screens

Service coverage

20+ services

Business shift

B2C to B2B white-label

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 10 // LISPINE-HEALTHCARE-WEBSITE/YEAR: 2024/ROLE: UI/UX DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

LISPINE Patient-Centered Website

The redesign served patients searching for back, neck and spine-care information, including older and anxious visitors. Mahad joined after multiple homepage revisions and helped recover the delayed project by completing 34 inner pages within an approved visual direction. He collaborated with designers and developers on responsive delivery and made controlled WordPress adjustments when late content changes threatened the schedule.

FigmaWordPressResponsive Web DesignHealthcare UXInformation ArchitectureUI Design

Read Case Study Live System

EMPIRICAL METRICS & EVIDENCE

Inner pages

Reported bounce rate

55% to 32%

Reported conversions

2×

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 11 // LEGAL-AI/YEAR: 2024/ROLE: AI APPLICATION DEVELOPER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Legal AI Contract Analyzer

Legal AI combines a Streamlit interface with a FastAPI backend and LLM-assisted document analysis. Users can upload supported contract documents, create retrievable chunks and ask questions about the uploaded material. The project demonstrates document ingestion, semantic retrieval and structured prompting, but it is not a substitute for legal advice and has no verified attorney-reviewed accuracy benchmark.

PythonStreamlitFastAPIGroqKimiChromaDBVector SearchDocument Parsing

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 12 // AURUS-BOOKS/YEAR: 2023–2024/ROLE: PRODUCT DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Aurus Books

Aurus Books was designed as an accounting and bookkeeping product for small and medium-sized businesses. Mahad rebuilt the information architecture after receiving the complete 50-page requirements document, then designed connected workflows for invoicing, expenses, clients, vendors, time tracking, CRM, reporting and a client dashboard. Competitive analysis informed familiar interaction patterns, but the product remained an MVP design and did not enter production.

FigmaProduct DesignUX ResearchCompetitive AnalysisWireframingInteractive PrototypingFintechInformation Architecture

Read Case Study

EMPIRICAL METRICS & EVIDENCE

Requirements

50 pages

DOC: VERIFIED SANITY RECORDEXPLORE →

Loading content...

SELECTED WORK // 2019–NOW[DOC: MMB-WRK-2025 // 12-COL MONOGRAPH]

SYS.SPEC: VERIFIED ARCHIVAL RECORDS·SYSTEM VERIFIED

Work That Moves From Interface to Intelligence.

A verified collection of AI products, full-stack applications, enterprise SaaS platforms, machine-learning projects, automation workflows, and product-design case studies.

AUDIT MANDATEVERIFIED 1.0

Each project identifies my role, what was actually completed, and the current maturity of the work. No theoretical mockups represented as live software.

PORTFOLIO_LATENCY: OPTIMIZEDCOORDINATES: 33.6844° N, 73.0479° E

DISPLAYING 12 VERIFIED SPECIMENS

SPECIMEN ID: 01 // AQL-ENTERPRISE-AI-WORKSPACE/YEAR: 2026/ROLE: AI PRODUCT DESIGNER AND PROTOTYPE DEVELOPER

[ACTIVE PRODUCT]

CATEGORY: AI Product

AQL Enterprise AI Workspace

Next.jsReactTypeScriptEnterprise UXRAG ArchitectureBGE RerankerGoogle DriveNotionJira

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 02 // DOCKET/YEAR: 2026/ROLE: FOUNDER AND PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Docket by ThinkSplice

MCPCloudflare D1Drizzle ORMTypeScriptVinextManaged AuthenticationJSON-RPCKanban UX

Read Case Study Live System

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 03 // MIMAR-STUDIOS-PLATFORM/YEAR: 2026/ROLE: FRONTEND AND PRODUCT DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

mimAR Studios Digital Platform

Next.jsReactTypeScriptTailwind CSSWebGLResponsive DesignAccessibilityPerformance Engineering

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Localized visual assets

Reference viewports

1440 px and 390 px

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 04 // TIDYOS/YEAR: 2026/ROLE: AI PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

TidyOS

PythonSQLiteSQLite FTS5LangGraphWatchdogDocument ExtractionOCRVector SearchDesktop Automation

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Automated tests

Implemented phases

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 05 // TRIPHISAAB/YEAR: 2026/ROLE: ANDROID AND AI PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

TripHisaab

KotlinJetpack ComposeAndroidRoomGroqNotification ListenerLocal-first ArchitectureStructured Expense Parsing

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 06 // VERDICT-AI/YEAR: 2026/ROLE: AI PRODUCT ENGINEER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Verdict AI

Next.jsReactTypeScriptSupabaseGroqStructured LLM OutputProduct Analytics UX

Read Case Study Repository

EMPIRICAL METRICS & EVIDENCE

Verified structural deliverable. Detailed metrics documented in full case study.

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 07 // CARDIOSCAN-AI/YEAR: 2026/ROLE: MACHINE LEARNING ENGINEER AND FULL-STACK DEVELOPER

[ACTIVE PRODUCT]

CATEGORY: AI Product

CardioScan AI

Next.jsReactTypeScriptTailwind CSSPythonFastAPITensorFlowKerasVGG16ResNet50V2DenseNet121SupabaseGroqHugging Face SpacesDocker

Read Case Study Live System Repository

EMPIRICAL METRICS & EVIDENCE

Original data

724 stress and 724 rest images

Augmented data

Approximately 2,000 paired samples

Model comparison

3 CNN backbones

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 08 // MAHAD-AI-PORTFOLIO/YEAR: 2026/ROLE: AI PRODUCT ENGINEER AND SYSTEM DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Mahad AI Product Engineering Portfolio

Next.jsReactTypeScriptTailwind CSSFastAPIPythonSanityNeon PostgreSQLQdrantmultilingual-e5-smallLangGraphONNX RuntimeMLflowLangSmithGroqSSEDockerElevenLabs

Read Case Study Live System Repository

EMPIRICAL METRICS & EVIDENCE

Router test accuracy

85%

Router p95 latency

16.77 ms

Assistant refusal score

100%

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 09 // BUSYFILE/YEAR: 2024–2026/ROLE: FOUNDING PRODUCT DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product

Busyfile

FigmaProduct StrategyProduct DiscoveryUser ResearchInformation ArchitectureDesign SystemsInteractive PrototypingFintechEnterprise SaaSMulti-tenant Product Design

Read Case Study

EMPIRICAL METRICS & EVIDENCE

Interface scope

500+ screens

Service coverage

20+ services

Business shift

B2C to B2B white-label

DOC: VERIFIED SANITY RECORDEXPLORE →

SPECIMEN ID: 10 // LISPINE-HEALTHCARE-WEBSITE/YEAR: 2024/ROLE: UI/UX DESIGNER

[ACTIVE PRODUCT]

CATEGORY: AI Product