Building at the speed of light — at the intersection of AI, legal, health, digital, education, and human-centered design.

AI-native legal workspace for law firms and in-house teams

Casengine turns matter management, document review, and due diligence into AI-assisted workflows — so lawyers can analyse hundreds of documents at once, codify firm standards as reusable playbooks, and get cited, grounded answers about case files in seconds. Bilingual (Arabic + English) with full RTL support and UAE-law-aware playbooks, and plugs into existing Practice Management Systems so it fits firm workflows instead of replacing them.

• Citation-backed Chat (RAG) — two-stage hybrid retrieval (semantic + keyword) with reranking pulls the exact clause out of hundreds of pages; every claim cites its source document and the system refuses rather than guess when the file doesn’t support an answer — cite-or-refuse, no hallucinated case law, in Arabic or English
• Document Review at scale (Tabular Review) — bulk-extract parties, dates, governing law, liability, and any custom data point across dozens of documents into one spreadsheet view, each cell carrying a confidence score and a citation back to the source clause — with templates for NDAs, employment, M&A, and service agreements, plus columns defined in plain language
• Playbooks — codify firm red-lines as reusable rule sets; the platform auto-detects each contract’s type, applies the matching playbook, and flags violations, missing clauses, and non-compliant terms by risk — UAE-law-aware (Civil Code, Labour Law, PDPL, Companies Law, RERA)
• Matters — case-centric workspaces where documents, extractions, playbook findings, and chat are all scoped per-tenant and per-matter

An AI-native operating system for academic accreditation

Accreditopia replaces the Google Drive + Word + Excel chaos that academic programs use to prepare for accreditation with a purpose-built platform organised around the standards programs are actually judged against. The beachhead is CAPTE — the body that accredits the ~680 Doctor of Physical Therapy and PTA programs in the United States — with a roadmap to every health-professions accreditor (OT, SLP, nursing, pharmacy). The goal: turn a once-every-five-years scramble into a continuous, low-anxiety operating practice.

• Standards Engine — a hand-audited backbone of all 91 CAPTE elements with the 2016→2024 crosswalk; every AI feature grounds against it
• Evidence + Curricular Library — auto-extraction parses each syllabus, CV, policy, or transcript once into two parallel stores: retrieval text and typed JSON
• Narrative Builder — per-element rich-text editor with a Citation Validator that blocks unsourced claims and a One-Voice Consistency Check across sections
• Outcome Intelligence — forward-looking risk on graduation rates, NPTE pass rates, and employment trends — flagging Standard 1C exposure two years before the on-site visit
• Edwin — the AI persona: Claude Sonnet 4.6 + Opus 4.7 on AWS Bedrock, six modes (Drafter, Mapper, Validator, Forecaster, Coach, Extractor), cite-or-refuse by construction
• Hardened RAG — hybrid vector + BM25 retrieval, per-tenant indexes, post-validated citations, PII redacted at extraction time

“Because ‘Looks Good To Me’ isn’t a legal defense.”

An autonomous AI auditor that lives in your GitLab workflow. Every merge request gets audited for copyleft license conflicts, leaked secrets, and PII — with findings grounded in a versioned library of nine professional legal briefs (Rule‑Cite). When a high-risk violation is found, LGTM opens a remediation MR with a compliant code swap and signed reasoning. Every decision becomes a defensible audit record.

• Dual-Retrieval RAG — public briefs + project precedents
• Rule‑Cite library — nine versioned legal briefs (RC‑01 to RC‑09)
• Tiered enforcement — HIGH blocks, MEDIUM flags, audit-trail always
• Agent advises, team decides — no auto-merges, ever

AI clinical copilot — ambient scribe and clinical workflow platform

Medwork.life turns doctor-patient conversations into structured medical documentation in real time. It listens to the visit, generates SOAP/BIRP notes, extracts actionable tasks (orders, referrals, follow-ups), and links every clinical statement to medical evidence — cutting documentation time by ~75% and freeing clinicians from the 2:1 paperwork-to-patient ratio that drives burnout. Beyond the visit, clinicians can ask the platform ad-hoc clinical questions and get streaming, citation-backed answers from a curated medical knowledge base, tuned to their country’s guidelines. EHR-agnostic via SMART on FHIR (Epic), HIPAA-aligned on AWS with a BAA in place, and multilingual across English, Swedish, French, and Arabic with 40+ specialties.

• Live AI Scribe — real-time hands-free transcription with speaker diarization, dialogue → SOAP/BIRP
• Smart Task Generation — orders, prescriptions, referrals, sick notes, each task traceable to the source utterance
• Evidence-connected notes (RAG) — clinical findings linked to citations from a medical knowledge base for full auditability
• Clinical Evidence Chat (RAG) — ask any clinical question and get a streaming, cited answer grounded in a curated knowledge base (PubMed, WHO, NICE, CDC, Cochrane, FDA labels). Answers are country-aware — the active guidelines adapt to the clinician’s region (US, UK, Canada/Québec, Sweden, France, Cyprus, Greece, Gulf), so a UK doctor sees NICE/BNF while a Swedish one sees Socialstyrelsen/FASS — with reliability-graded inline citations, cite-or-refuse grounding, in the clinician’s language
• Multi-agent AI — a state-machine router tracks clinical phase (History → Exam → Assessment → Plan) and dispatches to specialised agents (Note, Task, Evidence)
• EHR-agnostic — SMART on FHIR with JWKS, RS256-signed assertions; Epic-integrated, roadmap covers chart review & AI scheduling

AI voice assistant for small businesses — restaurants, clinics, salons, service shops

Alwaysdone.work gives small businesses their own always-on receptionist. Customers call a dedicated business number and an AI takes orders, books appointments, answers FAQs, and captures leads 24/7 — every conversation logged, transcribed, and synced to the dashboard. Built for the businesses that lose revenue every time a call goes unanswered. Onboarding is itself AI-driven: paste your business URL, the platform crawls it, extracts hours/menu/services, and auto-generates the assistant’s instructions and tool schemas — URL to live phone agent in minutes.

• AI Voice Assistant — natural, real-time phone conversations on a dedicated business number
• Order taking + appointment booking — menu-aware, calendar-aware, customer recognition, mishearing prevention
• RAG over your menus, price lists, and policies — upload PDFs/DOCX and the assistant’s knowledge updates itself
• Tool-calling architecture — the assistant doesn’t just talk; it executes real orders, bookings, and lookups via typed tool schemas
• Multi-tenant by design — per-business assistant, number, team, and Stripe-billed subscription, isolated via Supabase RLS

A weekly retrospective for your subreddit’s mod team

A first-party Reddit Devvit app that auto-publishes a newspaper-style retrospective inside a subreddit every Monday — one headline and fifteen behavioral sections covering mod workload, AutoMod accountability, decision contradictions, burnout signals, and team coverage gaps. Built solo in 28 days to fill the vacuum left when Moderator Toolbox was archived.

• 15 sections + an 18-candidate deterministic headline picker
• Closed-loop retrospective referencing last week’s findings
• Deep links into mod tools — one click to act
• ~5,200 LOC. Zero any. No LLM. No outbound network.