Data for Business. Data for Life.

Practical AI architecture
for teams that want delivery, not demoware.

We build the data pipelines, compliance platforms, and AI systems that the regulator, the audit team, and the engineers all trust. Fewer demos. More deliveries.

Book an audit What we do

Our Philosophy

Healthy systems share a shape.

Validate at the root. Build a single, well-formed trunk. The decisions and dashboards (the fruit) only stay accurate if the ground and the trunk are sound. We work the whole tree, not just the bit you can see.

Ground Up

Validate at Source The cheapest place to fix bad data is where it gets typed. We audit ingestion and lock the chain of custody before it has a chance to drift.

The Trunk

One Source of Truth We build the kernel that the rest of the business reads from, so reports, audits, and AI agents all see the same data.

The Fruit

Decisions That Hold Up Dashboards backed by audit trails, AI answers grounded in the source, and KPIs that survive the next regulator visit.

Diagnose before you prescribe.

Most rewrites we get called in to fix were prescriptions written before anyone diagnosed the problem. We borrow the medical workflow because it forces the habit. Diagnosis. Prescription. Surgery. Recovery. In that order.

See how we work

Full-Stack Logic

From data pipelines to web platforms. We build the systems your team will still understand in two years, and the regulator can audit on day one.

What we do

Critical Sectors

Built Environment, Energy, Workforce, and Commercial. Security-first, compliance-led, designed for people who get audited.

See the sectors

Research & Labs

The proving ground. Sovereign AI, local LLM orchestration, and the consumer apps we build to test our own architecture before clients pay for it.

Enter the lab

Featured Projects

CoreThread

A UK building-safety compliance platform built for the Building Safety Act 2022. CoreThread digitises the Golden Thread across the eight regulated pillars of a building, with offline-first mobile tools that work in the rain, sovereign per-client data, and a human-in-the-loop AI assistant that suggests but never decides.

https://corethread.io

https://dendro-logic.com/wp-content/uploads/2026/01/Icon-only.png

https://dendro-logic.com/wp-content/uploads/2026/01/2026-01-25-21_44_33-Antigravity-474x1024.png

https://dendro-logic.com/wp-content/uploads/2026/01/2026-01-25-21_46_44-Antigravity-474x1024.png

The Problem

The Building Safety Act 2022 expects every higher-risk building to carry a Golden Thread of information. An unbroken, auditable record of how it was designed, built, modified, inspected, and maintained. The regulator can ask for it at any point, and the evidence has to be produced in real time.

Most portfolios cannot answer that ask. Compliance lives in PDFs on shared drives, in chains of forwarded emails, in paper folders in site offices, and in the heads of long-serving staff. Site workers (surveyors, electricians, fire-door inspectors, gas engineers) are the people generating the data, but they are given desk-grade software that does not work where the work happens. On a roof. In a basement. In the rain. On patchy signal. So the data either does not get captured, or it gets captured twice, badly, weeks later.

What’s at Stake

For building owners, fragmented compliance is now financial risk. Insurance premiums climb. Lenders ask harder questions. Sales and refinancing slow when due-diligence packs cannot be produced quickly. For site teams, it means double-handling and dangerous gaps. For residents, it means the safety case for their home is held together by other people’s spreadsheets. The current pipeline does not pass the Building Safety Act test, and everyone knows it.

The Solution

CoreThread is a UK-only compliance platform built around an 8-Pillar Model: Fire, Electrical, Gas, Water, Structural, Lifts (LOLER), Hazards, and Energy (EPC). Each pillar has its own regulatory engine, its own evidence shape, and its own colour-coded UI, so a fire-door inspector sees a fire-door interface, not a generic form. Every event flows into a single, Merkle-verified Golden Thread that exports straight to BS 8644-1 XML.

Built for Site Reality, Not Office Reality

The mobile app is offline-first. WatermelonDB on-device storage. A 72-hour offline lease via signed token, so a surveyor can work a full week in a rural site with no signal and re-sync cleanly when they get back. Rain Mode swaps to a big-button UI with offline Whisper speech-to-text for data entry in bad weather. Accessibility is non-negotiable: high-contrast palette, dynamic type to 200 percent, designed for poor lighting and gloved hands. If the tool fights the worker, the data does not get captured.

AI That Suggests, Never Decides

Mimir is CoreThread’s AI assistant. It runs in two halves. Mimir Edge does on-device inference in under 100 milliseconds with no cloud calls, so a worker on a roof gets instant help on patchy signal. Mimir Cloud handles deeper reasoning when connectivity is back. Both are bound by a strict Human-in-the-Loop protocol. Above 0.9 confidence, Mimir auto-fills and the user validates. Between 0.7 and 0.9 it suggests. Below 0.7 it logs a quiet background flag, never a confident wrong answer. Every override requires a Reason Code.

Sovereign by Design

Each client gets a sovereign Spoke off a central Hub: separate VPC, separate database, separate storage. No cross-container joins. All data resides in eu-west-2 (London) with WORM-locked S3 evidence buckets. Account deletion is crypto-shredding: the encryption key is destroyed and the data becomes unreadable, the audit log keeps the receipt. Stolen-device wipe pushes a remote ACTION_WIPE that clears local SQLite and the keychain. Compliance is only worth what the data integrity behind it is worth.

Enterprise

CoreThread

Leap

Career mobility, built on proof. Leap turns real-world experience into verified, mathematically-validated skills. A portable Capability Ledger, proven through five-minute scenario challenges, guided by Opal the Empathetic Oracle, and trusted by recruiters and enterprises that need evidence, not keyword guesses.

https://leapcareer.io

https://dendro-logic.com/wp-content/uploads/2026/04/talent-og.png

https://dendro-logic.com/wp-content/uploads/2026/04/opal-og-1024x538.png

https://dendro-logic.com/wp-content/uploads/2026/04/enterprise-og-1024x538.png

The Problem

Hiring still runs on CVs and keyword-scanning software. If a person’s history doesn’t read like a tidy ladder, the scan misses them. Career changers moving industries. Parents returning after a break. Carers rejoining the workforce. Military veterans translating service into civilian capability. Gig-economy workers with messy, non-linear histories. All capable. All invisible to software that was never designed to see them.

The cost is paid by everyone. Recruiters fight over the same small pool of “safe” candidates and miss the bigger one sitting just outside it. Employers make multi-million-pound hiring decisions on proxies instead of proof. HR teams chase internal mobility targets without the data to act on them. The CV was never a good measure of capability. Keyword scans just made it worse.

The Cost of the Data Gap

Roughly half the working-age population doesn’t fit the linear-CV mould. Returners, switchers, veterans, gig workers, neurodivergent talent. Every one of them carries hard-won capability that the current pipeline cannot read. The result is a labour market that under-serves both sides: workers locked out of opportunities they could do tomorrow, and employers paying premium rates for the ones the scanner happened to surface.

The Solution

Leap replaces the CV-then-interview pipeline with a three-step flow: Translate, Prove, Match. Translate unstructured work history into the skills employers actually recognise, jargon-free, no HR vocabulary required. Prove them through five-minute capability challenges built around real-world scenarios, not coursework or paper certificates. Match directly to roles that need those specific skills, with the reasoning behind every score shown in plain English.

The verified results are written to a tamper-proof Capability Ledger that belongs to the user, not the platform. It exports as a portable Skills Passport that travels with them between jobs, sectors, and life stages. No-one has to re-prove themselves every time the role changes.

Three Portals, One Source of Truth

Talent is for individuals and families. Career exploration, the Capability Ledger, scenario challenges, and Opal-led guidance, in a tone built for real life rather than corporate HR. Recruiter is for agencies and direct hirers. A feed of “Transferable Gems”, anonymised by default, surfaced by the skills they actually have rather than the keywords they happen to use. Enterprise is for HR directors and people teams. Internal mobility, succession planning, retention modelling, and pay-equity enforcement, all running off the same verified evidence the talent side owns.

Opal, the Empathetic Oracle

Opal is Leap’s Synaptic Intelligence. She watches patterns in real work, offers low-friction verification prompts in the flow rather than yanking people out of it, and translates lived experience (parental leave, caregiving, military service, a winding career) into language employers understand. She also keeps an eye on live market demand and suggests learning paths with a clear return on investment, so users aren’t guessing where to put their next ten hours.

Built to Be Trusted

UK-sovereign data storage. ISO 27001 aligned. Cyber Essentials certified. Every match score carries a mandatory plain-English Display Reasoning module. The matching engine is regression-tested against a 10,000-career Ground Truth dataset and governed by an independent Data Integrity Board. Career mobility and hiring decisions should rest on the truth, and the truth has to be auditable.

Enterprise

Leap

Labs

Crystal Context

Crystal Context acts as a pre-processing layer for the AI-native developer, distilling verbose logs and messy thoughts into high-density prompts to maximise context window efficiency.

Cub and Compass – Infant Tracker

The ultimate offline command centre for modern parenting. From tracking the last dose of Calpol to generating wholesome recipes based on your baby's safe foods, Cub & Compass turns parenting data into confidence.

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260125-230505.png

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260126-061307-459x1024.png

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260126-061141-459x1024.png

Parenting is a blur of sleepless nights and critical questions that often go unanswered in the fog. “When exactly did we give the last dose of Ibuprofen?” or “Has it been three hours since the last feed?” aren’t just logistical queries—they are sources of genuine anxiety. Relying on memory when you are sleep-deprived is dangerous, and paper logs are just another thing to lose.

Then there is the “weaning trap.” It starts with enthusiasm but often degrades into a safe rotation of pouches and chicken nuggets because planning complex, safe meals for a baby is mentally exhausting. You end up stuck in a culinary rut, afraid to experiment because you can’t quite remember if they reacted to strawberries last week. You need a tool that doesn’t just record what happened, but actively helps you decide what happens next.

Cub & Compass – Infant Tracker is more than a tracker – it’s your intelligent co-pilot. It effortlessly manages the vital stats for your entire brood with multi-child profiles, tracking sleeps, feeds, and nappies with a fluid interface designed for one-handed use. The dedicated medication tracker answers the “safe to dose?” question instantly, providing a digital safety net for sick days. Plus, with secure Google Drive backups, your family’s data is safe forever without sacrificing privacy.

Where it truly shines is the Auto Weaning Recipe Expander. Instead of leaving you to guess, the app acts as a smart chef. As you log “safe” ingredients, it dynamically generates recipes that combine foods your baby already loves with exactly one new ingredient. This encourages a diverse, wholesome diet and gently pushes you away from processed safety foods, making weaning a journey of culinary discovery rather than a chore.

Labs

Cub and Compass – Infant Tracker

Sparky Mate

A robust digital toolkit designed to replace "guessing" with regulation-grade calculation. We took a basic utility concept and engineered it into a BS 7671 compliant companion for the modern electrician.

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260124-100258.png

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260124-100306-459x1024.png

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260124-100313-459x1024.png

Enhancing the Vision
Dunamis Electrical approached us with a clear goal: to digitize the “On-Site Guide” that every electrician carries in their van. Their original concept was a simple reference app, but at Dendro Logic, we saw an opportunity to do more than just display static tables. We wanted to build a tool that actively solved problems on the job site.
We expanded the scope from simple reference to active computation. Instead of just showing a Zs table, we built the calculator that validates it against your specific breaker type. We took the static concept of “Safe Zones” and turned it into an interactive visualizer. By injecting our logic-first approach, we transformed a passive reference manual into an active, essential partner for daily electrical operations.

Engineering Compliance
Reliability was our non-negotiable metric. In the electrical trade, a wrong calculation isn’t just a bug; it’s a safety hazard. We implemented rigorous testing suites to ensure every output—from Voltage Drop (VD) = 2 x L x R x I, to Adiabatic checks—matched BS 7671 regulations down to the decimal. We ensured these complex calculations run instantly and 100% offline, acknowledging that the most critical work often happens in basements without signal. We even included a conduit fill tool to work out what size / shape conduit is required for the proposed cable runs!

The result is “Sparky Mate,” a clean, distraction-free utility suite. We stripped away the ads and the fluff to focus purely on utility: Cable support spacing, SWA gland sizing, and ROI calculators for LED upgrades. It is a testament to how Dendro Logic can take a client’s domain expertise and wrap it in a robust, scalable technical framework.

Labs

Sparky Mate

The Decider

The Decider A beautifully tactile decision-making app that cures analysis paralysis. Feature-packed with 3D physics-based coins, wheels, and dice, plus a dedicated D&D mode and smart context recognition. Perfect for settling arguments, picking takeaways, or ruling the tabletop.

https://dendro-logic.com/wp-content/uploads/2026/01/IMG-20260120-WA0009.jpg

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260122-220722-459x1024.png

https://dendro-logic.com/wp-content/uploads/2026/01/Screenshot_20260122-220811-459x1024.png

The Problem

We’ve all been there: staring blankly at a list of takeaway delivery apps, having the circular “I don’t mind, what do you want?” conversation for the tenth time, or debating who should wash the dishes tonight. Decision paralysis is real, and it drains the fun out of your free time. Sometimes, you don’t need a logical pros-and-cons list; you just need a nudge. You need an unbiased third party to cut through the noise and make the call for you, instantly and definitively.

The Solution

Enter The Decider, an app that turns choice into a game. Whether you’re flipping a coin for dinner, spinning the wheel for a movie, or rolling a D20 for a critical hit, The Decider handles it with satisfying 3D physics and polish. Create custom lists like “Takeaway Battle” or “Weekend Activity,” and watch as our Smart Icon system automatically brings your options to life. With a built-in dedicated D&D mode, customizable dice materials, and a “Battle Mode” to pit specific options against each other, it’s the ultimate tool for settling disputes, gamifying chores, or just adding a spark of randomness to your day.

Labs

The Decider

View All Projects

Get in touch

Tell us what's broken, what you're building, or what you'd like to know. We reply to everything.

Name

Mobile Number

Sector

Message

Engineering

Long-form technical deep dives: benchmarks, architecture studies, and hands-on field notes.

May 06, 2026 / Engineering

DGX Spark Optimisation, Four Iterations On: From Defaults to 308 Tokens Per Second Sustained

Four iterations of inference tuning on a single DGX Spark and gpt-oss-120b. Same hardware, same model, 4.3x single-stream gain through flag tuning, an EAGLE3 negative result, a from-source vLLM build, and a one-line CUDA graph fix.

Read Deep Dive

Glowing crystalline brain at the centre of a chaotic data field, knowledge filaments converging from below

May 01, 2026 / Engineering

The AI Productivity Paradox, and the Team-Context Infrastructure That Fixed Mine

Written by Mike McGreal, Dendro Logic | Engineering | Published: 1 May 2026 The contradiction nobody is talking about loudly enough Adoption is real. Stack Overflow’s 2024 Developer Survey reports 63.2% of professional developers now use AI tools, with another 13.5% planning to adopt soon. 81% identify productivity as the biggest benefit they hope for. […]

Read Deep Dive

Concurrency race: single sequential stream far behind a dense bundle of parallel streams surging past the finish line

Apr 22, 2026 / Engineering

NVIDIA DGX Spark Concurrency Benchmark: 120x Throughput the Single-Stream Reviews Miss

Benchmarking the NVIDIA DGX Spark (GB10, 128GB) reveals 120x more throughput at concurrency 256 than single-stream reviews show. Full vLLM config and production workload results.

Read Deep Dive

View All Engineering

System Insights

Three abstract glowing clusters representing amnesia, hallucinated endpoints, and tribal silos in AI-augmented engineering.

May 05, 2026 / Engineering Reality

Amnesia, Hallucinated Endpoints, Tribal Silos: Why AI Made Your Team Busier, Not Faster

Key Takeaways A randomised trial by METR found AI made experienced developers 19% slower, while they believed it had sped them up by 20%. Adoption is high. Trust is low. Only 24% of developers trust AI-generated code “a lot” or “a great deal” (DORA, 2024). The bottleneck has moved from writing code to debugging code […]

Apr 27, 2026 / Agent Grounding

Your Team’s Slack Is the New Ground Truth

Key Takeaways Chat is where most team decisions are made and where most team context lives. Ignoring it is ignoring most of the team’s institutional memory. A scheduled job at end of day reads each conversation and produces a structured summary: decisions, blockers, progress, open questions, follow-ups. Each summary chunk carries a citation back to […]

Apr 24, 2026 / Agent Grounding

Decisions You Do Not Have to Relitigate: ADRs in the Agent Loop

Key Takeaways An ADR captures context, decision, alternatives considered, and consequences. The full reasoning, not just the outcome. ADRs are sequentially numbered, immutable once written, and superseded (not edited) when later decisions overrule them. The brain exposes a `record_adr` MCP tool that lets any agent session create a new ADR inline as the decision is […]

View All Insights

Practical AI architecture for teams that want delivery, not demoware.

Healthy systems share a shape.

Ground Up

The Trunk

The Fruit

Diagnose before you prescribe.

Full-Stack Logic

Critical Sectors

Research & Labs

Featured Projects

The Problem

What’s at Stake

The Solution

Built for Site Reality, Not Office Reality

AI That Suggests, Never Decides

Sovereign by Design

CoreThread

The Problem

The Cost of the Data Gap

The Solution

Three Portals, One Source of Truth

Opal, the Empathetic Oracle

Built to Be Trusted

Leap

Crystal Context

Cub and Compass – Infant Tracker

Sparky Mate

The Problem

The Solution

The Decider

Get in touch

Engineering

DGX Spark Optimisation, Four Iterations On: From Defaults to 308 Tokens Per Second Sustained

The AI Productivity Paradox, and the Team-Context Infrastructure That Fixed Mine

NVIDIA DGX Spark Concurrency Benchmark: 120x Throughput the Single-Stream Reviews Miss

System Insights

Amnesia, Hallucinated Endpoints, Tribal Silos: Why AI Made Your Team Busier, Not Faster

Your Team’s Slack Is the New Ground Truth

Decisions You Do Not Have to Relitigate: ADRs in the Agent Loop

Project Gallery

Practical AI architecture
for teams that want delivery, not demoware.