Adel Labs · Design, engineering & applied AI
Commission a brand, website or app; turn technical work into interactive tools; or equip your team with AI workflows built around its documents and tasks. From business dashboards to research workspaces, each project is developed around how you will use it.
Keep confidential research on your own hardware with a private AI workspace, optional unrestricted models configured to reduce refusals, and offline operation tested before handover.
Defined scope · Source files and operating notes.
Digital Products
Research, design and engineering, with AI where it helps.
Drag horizontally or use the left and right arrow keys to rotate. Home resets the view. Select a capability to highlight its contribution.
Selected work · Five disciplines
Open a working example, inspect its recording or explore the complete collection.
Explore all 32 projects →01 / Collection
Brand identities, company websites and digital storefronts.
View collection · 7 projects →4 selected projects
Brand, website & social content · Client work
A coordinated website and social content system for a UAE hosiery and shapewear brand, extending its established identity across product pages, campaign imagery, Instagram posts and carousels, reels and outfit edits.
Website · Infrastructure studio
A website for an infrastructure studio, with a 3D architecture scene and a lab of engineering field notes.
Client work · Art & culture
A trilingual review website for the foundation preserving the legacy of Mohamed Melehi (1936–2020), connecting biography, chronology, bibliography, archives and exhibitions.
Client work · Personal brand
A trilingual French–English–Arabic hub for a Marrakech naturopath and certified coach.
02 / Collection
Personal applications for documents, planning and business decisions.
View collection · 8 projects →4 selected projects
SAMPLE demo · Business overview
Explore a SAMPLE service business: cash scenarios, invoice payments, client stages, weekly capacity and validated CSV/JSON import and export.
Personal product · PDF Text Studio
Change or add text on a PDF or scanned page. Import files, recognise text in English, French or Arabic, review edits and export at print resolution.
Personal product · Markets Terminal
Scheduled exchange sessions, provider charts, watchlists, economic and earnings calendars, and news in one browser workspace.
Personal product · Daily planning
A personal dashboard for the clock, local conditions, focus tasks, market readings and daily activities, with four navigable views.
03 / Collection
Explore chemical and gas EOR through interactive studies, live scenario calculations and source publications. Adjust model parameters to compare outputs, inspect the assumptions, and follow the supporting manuscripts, data workbooks and presentations available with each study.
View collection · 9 projects →4 selected projects
Program hub
Explore surfactant recovery in carbonate reservoirs through capillary desaturation, retention screening and economics. Compare live scenarios, then inspect the supporting manuscripts, data workbooks and program presentations.
Institutional research · Khalifa University / ADNOC
Program hub
Explore gas recovery in unconventional reservoirs through miscibility, operating pressure, gas composition, foam and hybrid recovery. Adjust scenario inputs, compare calculated outputs and follow the published papers behind each study.
Institutional research · Texas A&M University
Study 03
Discounted-cash-flow screening that turns surfactant retention into net present value, return rate, unit technical cost and breakeven retention, on twelve corefloods, with adjustable assumptions.
Institutional research · Khalifa University / ADNOC
Report 01
Slim-tube MMP across six CO2–crude systems, pressure-dependent huff-n-puff recovery, gravity-stable displacement, and an operating-pressure planner.
Institutional research · Texas A&M University
04 / Collection
Teaching companions with live calculations and worked explanations.
View collection · 4 projects →4 selected projects
Learning resource · Teaching companion
Ten interactive modules combine published PVT correlations with labelled schematic explanations. Change inputs, inspect the result and review the stated basis.
Learning resource · Teaching companion
Twelve modules on the mechanisms that mobilise trapped oil: capillary number, fractional flow, sweep, polymer, gas and surfactant processes, computed live.
Learning resource · Teaching companion
From log response to net-pay and volume estimates. Explore invasion, resistivity, porosity and saturation with visible assumptions and worked calculations.
Learning resource · Teaching companion
Material balance, aquifer influx and displacement: thirteen reservoir-engineering calculations, each computed live beside its explanation and a visual model.
05 / Collection
Technical workbenches with visible inputs, units and exports.
View collection · 4 projects →4 selected projects
SAMPLE workbench · Coreflood experiments
Plan and analyse laboratory coreflood experiments: geometry, permeability, injection, production and capillary desaturation. Editable SAMPLE data and local exports.
Engineering tool · Calculator
Fitted retention–recovery relation, chemical demand and unit technical cost, an NPV surface over oil price and retention, and the breakeven envelope.
SAMPLE workbench · Phase behaviour
A ten-vial phase scan with measured IFT, corrected static adsorption and dynamic retention balances.
Engineering tool · Calculator
Compare trapping numbers on interstitial and Darcy bases, screen field rates against measured onsets, fit desaturation data and convert petroleum units.
What you can commission
From a brand and website to a configured development environment or a private research workspace. Commission the complete setup, with reusable workflows and the knowledge to operate it.
01 / Brand, websites & apps
Commission your brand identity, website or personal app. Connect product information and clear navigation to the task your customers or colleagues need to complete.
Brand assets, responsive pages or a working app, editable source and a walkthrough.
02 / Agent setup & workflows
Get Claude Code or Codex configured with specialist agents, reusable skills, project context and review steps. Repeat research, design and development tasks from a prepared setup.
A configured environment, agent roles, skill library, workflow files and representative task checks.
03 / Private AI & workspaces
Search confidential documents and prepare cited drafts on your workstation or private server. Choose unrestricted research configurations when broader response behavior is needed.
A model workspace matched to your hardware, access settings, documented data routes and tested workflows.
04 / Research & engineering
Connect measurements, assumptions and calculations in a report, dashboard or technical tool. Package suitable workflows for offline use with local imports and exports.
Visible inputs and assumptions, calculation checks, local import and export, and editable source.
Agents, skills & working methods
Equip your team with AI agents for marketing, coding, design and research, configured around your brand, documents and working methods. Reusable skills provide task-specific procedures, templates and checks. For larger assignments, coordinated agents divide the work, review results and prepare a connected handover.
Brand positioning and voice; search audits and content briefs.
Debugging, interface implementation and site checks.
Art direction, interface design, campaign and product content.
Source analysis, reference verification and numeric audits.
Technical writing, workbooks, presentations and interactive reports.
Agentic AI carries out a sequence of tasks with tools and checks along the way. A skill supplies a reusable method; an agent applies it; a coordinated team divides a larger assignment between specialist roles.
Each setup uses a selected skill set, tested on representative tasks. Setup, testing and handover are included in the agreed scope; continuing maintenance is scoped separately. Connected development tools and fully offline model workspaces are separate configurations.
A configured starting point
The value continues after the first task: project context is organised, recurring procedures become skills, and specialist roles have defined review points. The setup is refined against your work, so you can reuse it and extend it.
Your hardware · Your data
Standalone apps and local model workflows can run on your equipment with no cloud processing or runtime connection for the agreed task. Local storage, access, backup and exports are part of the scope; the complete workflow is tested disconnected.
Provider-backed Claude Code and Codex configurations use connected model services. A fully offline model workspace is a separate configuration.
Private AI · Local processing & research
Ask questions across your research library, trace answers to source documents, compare findings and prepare editable drafts in a workspace configured for your hardware. Model choice, document retrieval and offline operation are tested against your tasks before handover.
Optional unrestricted models are configured to reduce refusals. Source review and answer-quality checks remain part of the workflow.
Organise the documents and references agreed for your workspace.
Scope source-linked answers and comparisons across the selected material.
Review findings and drafts against their sources before using them.
Define the document formats, source trail and export route at handover.
These are commissioned capabilities, configured and tested for the agreed task.
Unrestricted model implementation
Assess quality, response time and operating cost on representative work.
Community variant
A candidate for document work, research and coding where reduced refusal behaviour matters. Validate the configuration against your tasks.
Implemented with Locus
Private deployment with an optional abliterated configuration. The Locus implementation is complete.
Private AI · Unrestricted model implementation
Explore sensitive research questions, compare arguments and work with private documents in an environment you control.
An abliterated model is an open model whose internal weights have been modified to reduce learned refusal behaviour. It can answer more of the research questions that a standard model may decline.
We configure the model, your document workflow and the interface on suitable local hardware. Fully offline operation is available after the complete setup passes a disconnected test.
Fewer refusals do not add knowledge or guarantee accuracy. We assess answer quality alongside refusal behaviour using tasks that matter to your work.
Qwen3.8-27B
AdvBench refusal results reported by the variant author
Variant-author results using greedy decoding and keyword scoring. On HarmBench, non-thinking mode refused 45 of 393 prompts (11.45%); thinking mode was reported at 0%. Adel Labs did not run these benchmarks. Refusal rate is not an answer-quality measure.
Read the variant & test results ↗ (opens in a new tab)A candidate for research, document work and coding. We test the selected configuration on Locus hardware before recommending it for your workload.
DeepSeek V4.1-Flash
DSpark speculative decoding
An abliterated configuration can be scoped for research workflows that need fewer refusals. The implementation status is stated above.
DeepSeek V4.1-Flash includes DSpark speculative decoding. The referenced abliterated FP8 option preserves the bundled DSpark head.
We size the selected checkpoint, context length and number of users around your workload, then test the model with your documents and tasks.
Deployment decision
Choosing the right architecture
Compare data handling, expected usage and the work of operating each setup. Local, cloud and hybrid each fit different requirements.
| The decision | Local / on-premisesOwned or leased hardware | Cloud / APIHosted models and compute | HybridSplit by task and data |
|---|---|---|---|
| Data & offline access | Keep processing on site. Fully offline operation needs the entire workflow to run locally. | Requires connectivity and agreement on provider location, retention and access. | Keep sensitive tasks local. Define exactly what may leave for external processing. |
| Cost & workload | Assess against steady use; lease or hardware, energy and support costs continue when idle. | Usage charges can suit occasional work and bursts; compare costs at your expected volume. | Budget for local capacity, external usage and the work of connecting both. |
| Models & response time | Choose compatible open models within memory and concurrency limits. Benchmark your actual tasks. | Access hosted models; account for rate limits, network delay and provider changes. | Select models by task. Test quality, response time and the offline fallback. |
| Operations & ownership | Assign responsibility for updates, access, backups and hardware maintenance. | Provider-managed infrastructure; your team still manages access, integration and output review. | Maintain both environments, routing rules and recovery when a service is unavailable. |
We compare representative tasks, answer quality, response time, data requirements and total operating cost before recommending a deployment.
Adel Labs × Locus
A private model environment without buying the GPUs upfront. Locus provides the leased infrastructure; Adel Labs builds the working system around your research and team.
Locus offers dual NVIDIA RTX PRO 6000 Blackwell servers installed in the client’s environment, with configuration and maintenance scoped to the workload.
We install the selected model, connect approved documents and tools, and test the agreed tasks. Your handover includes the interface, configuration and operating instructions.
From leased hardware to a research environment your team can operate.
Keep model inference, document retrieval and outputs inside your environment. For offline delivery, we test the complete workflow with the external connection disabled.
Capacity and pricing follow an assessment of the model, document length, concurrent users and support needs. Any remotely hosted option is scoped separately.
Compare local, cloud and hybrid →Your data. Your way of working.
Turn your personal database, company records and documents into a workspace that answers questions with relevant source material.
Prepare a workspace brief ↗01 / Connected knowledge
Search approved documents and databases, retrieve the relevant passages, and review the sources behind an answer. Update the source material without retraining the model.
02 / Model adaptation
When a task needs it, fine-tune a compatible model on approved examples from your own work. Define the training data, target behaviour and evaluation before training starts.
Begin with your data, a task and examples of a useful answer. We assess the approach, access controls and deployment with you.
Project delivery
Scope, fees and timing follow the content, integrations, data handling and acceptance work involved.
Users, content, constraints and acceptance checks. A defined brief and delivery scope.
Review the structure with representative content. A working prototype you can assess.
Check behaviour, boundaries and outputs. The agreed build and recorded checks.
Review normal operation and responsibilities. Agreed source files, configuration, operating notes and a walkthrough.
Maintenance, updates and further development are scoped separately. Responsibilities are agreed before handover.
Prepare your project brief →
Founder-led delivery
Senior Research Engineer · Digital product development
My research at Khalifa University within the ADNOC program centres on surfactant recovery in carbonate reservoirs. At Adel Labs, I bring that attention to evidence into websites, private workspaces and practical tools.
You review the work with the person building it, from the first scoped task through handover.
Before we begin
Adel Labs builds brands, websites, apps and interactive research tools, and configures agent workflows and private model workspaces. Start with what you need people to do; we define the deliverables and how the result will be checked.
You receive a proposal covering deliverables, fee, schedule and review points before work begins. If a technical question needs testing first, we can scope a prototype before committing to the full build.
We agree where your files are stored and which systems process them before setup. The workspace can use your own hardware, a private server or agreed external services. Any offline workflow is tested without an internet connection before handover.
Compare operating modes →The proposal identifies the source files, configuration, documents and access you receive. Third-party model and software licences remain applicable. You receive operating instructions and an agreed backup/export route.
Handover includes the agreed checks, operating instructions and a walkthrough. Ongoing support, maintenance, update testing and further development are scoped separately.
Your next project
Start with the task, who will use the result and what needs to improve.
