Home › Azure Cognitive Services
Platform Engineering · Capability
Azure Cognitive Services.
For teams already on Azure, Cognitive Services give you speech, vision, and language APIs. We wire them into bespoke platform architectures where the rest of the stack is open-source.
Scope
What we do
- Integrate Azure Speech-to-Text and Text-to-Speech into production pipelines.
- Use Azure Form Recognizer behind a NATS-driven ingest queue.
- Add Azure Content Safety as an Envoy external-auth gate.
- Cache and cost-govern cognitive API calls.
Practical
Exercises we run
Small, repeatable drills we use on engagements and teach in workshops. Each has a lab setup, step-by-step outline, and measurable output.
Cost model
Approximate UK-South pricing for the services we reach for
Budgeting signal, not a quote. Azure publishes SKU-level pricing that changes quarterly; these ranges cover the usual-case tier + a premium tier we actually deploy, and point at where the real money goes once a workload is past the trial phase.
| Project | API family | Approx. £ per 1,000 requests (UK-South) | Where the real cost goes |
|---|---|---|---|
| Speech-to-Text | Standard real-time & batch transcription; also fast-transcription, custom speech endpoints, speaker diarisation. | ≈ £0.80 – £2.00 per 1,000 seconds of audio (S0 tier vs custom-endpoint premium). Conversation Transcription is priced per participant-minute separately. | Long calls and re-transcription. We always cache the final transcript on SeaweedFS keyed by audio hash; re-runs cost nothing. Custom models have a separate hosting charge whether you use them or not. |
| Content Safety | Text + Image moderation; Groundedness detection; Prompt Shields for LLM input filtering. | ≈ £0.30 – £1.00 per 1,000 text records. Image analysis is priced per image. Prompt Shields are often bundled with Azure OpenAI SKUs — check before double-billing. | High-traffic moderation gates. We run Content Safety as an Envoy external-auth filter with a short-TTL cache on exact-match inputs; the cache hit rate on repeat user text is usually 20-40%. |
| Language / Text Analytics | Sentiment, key-phrase, named-entity recognition, PII detection, summarisation, question-answering. | ≈ £0.50 – £1.50 per 1,000 text records for standard features. Question-answering is metered separately on provisioned throughput units. | Document-heavy workloads. Batching (up to 10 records per request) and embedding-then-filtering before hitting the API cut cost by 2–5× in practice. PII detection at the storage layer is strictly cheaper than at query time. |
Prices are indicative ranges observed on UK-South engagements in 2026 Q1; Azure moves SKUs frequently — always check the linked Microsoft page before budgeting. Volume discounts (Enterprise Agreement, Reserved Capacity) and regional routing can swing the effective rate by 30%+.
Further reading
More on Azure Cognitive.
Workshops we teach + field notes we're writing, all linked back to what you just read. See all workshops → See all field notes →
Hands-on: Azure Cognitive in a bespoke platform — 1-day workshop
Wire STT into an Envoy ext_authz edge pipeline with a SeaweedFS cache. Add a Content Safety gate. Measure cost savings honestly.
Scheduling soon →
Wire Speech-to-Text into an edge pipeline
Envoy `ext_authz` → sha256-keyed SeaweedFS cache → Azure STT on miss. Measured 25-30% Azure cost cut on 30%-repeat workload.
Draft →
Content Safety as a middleware gate
Per-route opt-in Envoy gate → Azure Content Safety. 500-item corpus eval + fail-open vs fail-closed decision record.
Draft →
Engagement
Hands-on: Azure Cognitive in a bespoke platform — 1-day workshop
Packaged engagement — we scope, build, and hand over with runbooks, against a specific SLA. Add to cart to request delivery; no price is billed up-front.
Neux Ltd
AI Infrastructure · Platform Engineering · London.
Since 2014.
Contact
Legal
© 2014–2026 Neux Ltd
Registered in England & Wales.