01 :: COMPANY_PROFILE
MOTD: "There is no cloud. It's just someone else's triplestore."
TENGU SOLUTIONS LTD is an independent IT consultancy specialising in semantic web technology, enterprise knowledge graphs, machine learning, and applied artificial intelligence.
We design and operate RDF-native data platforms: ontology-driven models (OWL, SHACL), SPARQL query layers, and graph storage on enterprise triplestores. On top of the graph we build the intelligence layer — retrieval-augmented and agentic AI systems that traverse the data model and explain their answers — and the classical ML pipelines that keep production processes measurable and optimised.
Ten+ years of production deployments in regulated environments: pharmaceutical manufacturing (GxP), the public sector, supply chain. We take systems from proof-of-concept to production. No slideware.
DOMAINS: pharma manufacturing (GxP) / public sector / supply chain
GRAPH: Anzo / Amazon Neptune / Stardog / GraphDB / Jena / Neo4j
AI: Bedrock / Azure OpenAI / frontier LLMs / RAG / agents / MCP
CLOUD: AWS / Azure / Databricks
MODE: PoC → production // direct B2B or via partner consultancies
02 :: CORE_SERVICES
SVC-01 :: KNOWLEDGE GRAPH ARCHITECTURE
Ontology engineering and enterprise knowledge graphs, end to end: domain modelling in OWL/RDF, SHACL validation as schema-as-code, SPARQL query engineering and performance tuning, deployment on enterprise triplestores. Including graph frontends that let non-technical users interrogate the graph — zero SPARQL required.
deps: RDF / RDFS / OWL 2 · SHACL · SPARQL 1.1 · SKOS / JSON-LD · Anzo · Neptune · Stardog · GraphDB · Jena · Neo4j · RDFLib
SVC-02 :: AGENTIC AI & RETRIEVAL SYSTEMS
LLM systems grounded in your data, not vibes: retrieval-augmented generation over your documents and your graph, agentic workflows orchestrating classification and enrichment, and agent-to-data access patterns so models can traverse the data model and explain outcomes.
deps: AWS Bedrock · Azure OpenAI · frontier LLM APIs · enterprise search & retrieval · vector stores · MCP & tool-use · eval harnesses
SVC-03 :: ML PIPELINE ENGINEERING
Classical machine learning where it pays for itself: multivariate process monitoring, gradient-boosted and ensemble prediction, simulation for process and supply-chain optimisation — engineered as maintainable, monitored pipelines, not orphaned notebooks.
deps: Python · scikit-learn · XGBoost / LightGBM · PyTorch & deep learning · Databricks / Spark / Hadoop · MLflow · Monte Carlo & simulation · statistical process control
SVC-04 :: PRODUCTION DATA PLATFORMS
The unglamorous part, done properly: ingestion pipelines, serverless architectures, CI/CD, infrastructure-as-code, SSO/identity integration, automated testing — plus the auditability and traceability that regulated (GxP) environments demand. Comfortable below the cloud line, too: Hadoop clusters, GPU systems, and enterprise API gateways kept running 24/7.
deps: AWS · Azure · Terraform · Docker / Kubernetes · GitHub Actions · serverless & event-driven · FastAPI / TypeScript · API gateways & B2B (XML / XSLT / WSDL) · on-prem Hadoop / GPU · observability & audit trails
03 :: SYSTEM_LOGS
// production milestones. filtering: MAJOR_RELEASES only — minor commits hidden. timestamps: ████ (NDA).
>> SYS.01 // GLOBAL_PHARMA_MANUFACTURING (GxP)
[████-Q█] v1.0 :: baseline RDF ontology deployed across isolated manufacturing sites.
[████-Q█] v1.2 :: laboratory information (LIMS) data integrated into the core graph.
[████-Q█] v2.0 :: formal GxP compliance audit passed. automated regulatory exports live.
[████-Q█] v2.5 :: zero-SPARQL search UI shipped. adopted on the manufacturing floor, no training required.
[████-Q█] v3.0 :: agent overlay merged. real-time batch-deviation querying active.
STATUS: [ ACTIVE_MAINTENANCE // SCALING_TO_NEW_SITES ]
>> SYS.02 // CONTINUOUS_MFG + SUPPLY_CHAIN
[████-Q█] v1.0 :: gradient-boosted yield-prediction pipeline deployed.
[████-Q█] v1.5 :: multivariate statistical process monitoring integrated.
[████-Q█] v2.0 :: closed-loop simulation live. drift detection caught anomaly #4092 (saved £££).
[████-Q█] v2.2 :: models migrated to managed pipelines. CI/CD fully automated.
[████-Q█] v3.0 :: long-run review confirms sustained double-digit yield uplift.
STATUS: [ VALUE_REALISED // PIPELINES_HANDED_OVER ]
>> SYS.03 // PUBLIC_SECTOR_████████
[████-Q█] v0.x :: architecture approved. infrastructure-as-code base layer deployed.
[████-Q█] v1.0 :: automated validation pipeline live for a regulated document corpus.
[████-Q█] v1.x :: agentic classification & semantic tagging online. ██████████████
STATUS: [ ██████ // DETAILS_WITHHELD ]
04 :: CONTACT
> INITIATE_TRANSMISSION
[ TRANSMIT >> contact@tengusolutions.com ]
PROTO: SMTP preferred. No fax. No carrier pigeon.
OPS: United Kingdom + Europe · packets accepted from any origin
MODES: direct B2B contract · specialist delivery partner to consultancies
RESPONSE: < 48h on business days · hand-pulled, cask-conditioned
// briefs, RFPs and awkward data problems welcome. attachments in open formats preferred.
EOF :: TRANSMISSION_FOOTER
YOU ARE VISITOR NO.
00133742
counter provided by TenguStats™ v0.9b — reloading will not inflate your importance.
Best experienced at 800×600 in Netscape Navigator 4.7.
Degrades gracefully to modern browsers. Hand-coded HTML — no frameworks,
no cookies, no trackers, and an irresponsible amount of JavaScript:
29 undocumented features ship in this build — the tengu head alone
has six ways to leave, and the shell runs deeper than its help admits.
Happy hunting — are you watching closely?
[ BACK_TO_TOP ^ ]
EOF — nothing ever ends.
Any resemblance to films, shows or franchises is affectionate parody by a developer having fun. All names, characters and properties belong to their respective megacorporations — no affiliation or endorsement implied.