# Atom Technologies > Dedicated software teams from Latin America, working US hours: a managed team run by our tech lead, or senior developers who join your team. Published prices and terms. - Managed software teams: Starter $29,500, Core $44,500, Plus $67,500 a month, full-time, run by our tech lead; 3-month initial term. - Developers in your team (staff augmentation): mid $7,000, senior $9,500, data/AI $11,500, lead $12,500, QA automation $6,500 a month; 10-business-day trial. - Blueprint: two weeks, $7,500 fixed: a plain-language scope, architecture, milestone plan, risks and a fixed monthly quote, yours to keep. - From a first call to a team at work in 4-6 weeks; a start 10-15 business days after signature; 30 days' notice; free replacement. - Atom Technologies is a trading name of Atom Digital Technologies LLP (England & Wales, OC450214); Panama City, on US Eastern time. ## Pages - [Dedicated software teams from Latin America](https://atomtechnologies.ai/): A dedicated team run by our tech lead, or senior developers in your team. Latin America, US hours, published prices and terms. - [Managed software teams](https://atomtechnologies.ai/teams/): A dedicated development team that works only for you, run by our tech lead: Starter, Core and Plus. - [Staff augmentation: senior developers](https://atomtechnologies.ai/developers/): Senior Latin American developers who join your team and work US hours, with a 10-business-day trial. - [Blueprint: scope, plan and a fixed monthly quote](https://atomtechnologies.ai/blueprint/): Two weeks, $7,500 fixed, for companies without a written scope. - [Pricing](https://atomtechnologies.ai/pricing/): Published monthly prices for managed teams and developers, and every term. - [How it works](https://atomtechnologies.ai/how-it-works/): From a 30-minute call to a dedicated team at work. - [For software companies](https://atomtechnologies.ai/for/software-companies/): Developers in your team, then managed teams. - [For companies that run on software](https://atomtechnologies.ai/for/companies-that-run-on-software/): Staffing and BPO, freight and 3PL, insurance and property management: the Blueprint first, then a team. - [Staff augmentation vs a recruiter](https://atomtechnologies.ai/not-a-recruiter/): Recruiters, employers of record and our two models, compared. - [FAQ](https://atomtechnologies.ai/faq/): Time zones, minimum term, trial, replacements, IP assignment, security and contracts. - [Company facts](https://atomtechnologies.ai/company/): Who we are, where we are, how an engagement starts and how we hire. - [Security and data handling](https://atomtechnologies.ai/security/): What we access, how we protect it and who owns the result. - [Contact](https://atomtechnologies.ai/contact/): Ways to reach us. - [Book a call](https://atomtechnologies.ai/book/): A 30-minute call with the Atom Technologies engineering lead. - [The managed team sheet](https://atomtechnologies.ai/free/team-sheet/): Free: how a managed team and developers in your team work, roles, prices and terms, on one page. ## Articles - [Architecture assessment: what would justify a rewrite?](https://atomtechnologies.ai/insights/architecture-assessment-repair-or-rewrite/): Turn a repair-or-rewrite debate into a bounded investigation. Define the evidence that would change the decision before commissioning another system. - [Backfill reconciliation: compare old and new data while writes continue](https://atomtechnologies.ai/insights/backfill-reconciliation-concurrent-writes/): A finished batch does not prove that a derived value is current. Use a PostgreSQL unit-conversion example to compare the same record version, handle concurrent changes, and account for every unresolved row. - [Capacity planning: define a completed operation before setting throughput](https://atomtechnologies.ai/insights/capacity-planning-completed-operation/): A request accepted, retried, or queued is not necessarily completed work. Use an asynchronous export example to define throughput, reconcile unfinished work, and set a testable capacity target. - [Cost per completed operation: include the work that did not finish](https://atomtechnologies.ai/insights/cost-per-completed-operation/): Compare service costs using attributable capacity, shared expenses, and all billable attempts. A worked example shows how retries can outweigh a lower per-attempt rate. - [Database restore rehearsal: what must pass before traffic returns?](https://atomtechnologies.ai/insights/database-restore-rehearsal-traffic-readiness/): A recovered database still needs a verified recovery point, compatible application dependencies, and operator evidence. Define those conditions before timing a restore rehearsal. - [Deployment health: prove a business operation before moving traffic](https://atomtechnologies.ai/insights/deployment-health-business-operation/): A passing health endpoint can miss a broken workflow. Define a bounded release probe that follows one accepted operation to its result and leaves enough evidence to investigate uncertainty. - [Durable transaction recovery after a lost response](https://atomtechnologies.ai/insights/durable-transfer-recovery/): A Rust reference study of durable receipt recovery when a transfer commits but its response is lost. Inspect the failure experiments and operating limits. - [External API contract drift: is this response still usable?](https://atomtechnologies.ai/insights/external-api-contract-drift-response-usability/): Decide whether a provider response has the right identity, meaning, version, and observation age. Then give the live contract check an owner who can act on a failure. - [Fallback-served LLM requests at release: separate passes, cost per task, and rate-limit headroom](https://atomtechnologies.ai/insights/fallback-served-llm-requests-release-criteria/): Before approving a production LLM feature, count fallback-served requests separately from primary-model passes. The release record should show the fallback step, cost per contract-satisfying task for each path, and rate- - [Choosing the first strangler slice off a legacy PostgreSQL core: plain read, locking read, or write](https://atomtechnologies.ai/insights/first-strangler-slice-postgresql-lock-and-visibility-evidence/): Pick a plain read slice when it delivers the learning you need, a locking read only when row-lock behavior is the learning and its evidence is in hand, and a write only when writes are required. Route production traffic - [Release gate for automatic LLM fallback on side-effecting AI calls](https://atomtechnologies.ai/insights/llm-fallback-release-gate-side-effecting-calls/): A proposed release gate: let read-only model calls fall back automatically, but for calls that trigger side effects, require an idempotency key, a reconciliation path, and serving-step evidence before enabling fallback. - [Load-test validity: when does a result support production capacity?](https://atomtechnologies.ai/insights/load-test-validity-production-capacity/): A request rate is meaningful only with its workload and environment. Use a catalog-cache example to examine data distribution, arrival patterns, shared resources, and overload behavior before relying on a load test. - [Message-schema evolution: who must understand a new enum value?](https://atomtechnologies.ai/insights/message-schema-evolution-new-enum-consumers/): A message can decode successfully while causing the wrong business action. Use an export-result enum to identify affected consumers, prepare mixed versions, and decide when new values may be emitted. - [Shedding order for an overloaded SaaS platform: the pre-incident test that settles it](https://atomtechnologies.ai/insights/overload-shedding-order-pre-incident-test/): The Google SRE Book supports per-customer limits and degraded responses as deliberate shedding mechanisms; this article proposes a controlled overload test that compares them with arrival-order errors against per-class S - [PgBouncer rolling restarts: why idle clients delay exit](https://atomtechnologies.ai/insights/pgbouncer-rolling-restart-client-lifetime/): A local PgBouncer 1.26.0 study shows why a committed transaction can leave the process waiting for an idle client to disconnect. - [PostgreSQL lock waits: assess an additive migration's blocking risk](https://atomtechnologies.ai/insights/postgresql-lock-waits-additive-migration/): Adding a column can avoid a table rewrite and still interrupt traffic. Trace the lock queue, bound the migration transaction, and define the evidence needed to approve the change. - [PostgreSQL migration rollback: find the first incompatible write](https://atomtechnologies.ai/insights/postgresql-migration-rollback/): An additive schema change can still make rollback unsafe. Use a scheduled-job example to plan compatible releases and test recovery after new data arrives. - [Production AI evaluation: which failures should block a release?](https://atomtechnologies.ai/insights/production-ai-evaluation-release-criteria/): A better average score can conceal a broken requirement. Define release criteria for a document-grounded assistant, its graders, and the evidence behind approval. - [Production AI model fallback: preserve the application's decision](https://atomtechnologies.ai/insights/production-ai-model-fallback-output-contract/): A backup model can return valid JSON and still change what the application does. Define semantic compatibility, abstention, and a shared latency budget before enabling provider fallback. - [Rust async cancellation at a database transaction boundary](https://atomtechnologies.ai/insights/rust-async-cancellation-database-transaction/): Separate the lifetime of a Tokio task from the outcome of its PostgreSQL transaction. Give one resource owner responsibility for the commit decision, bounded cleanup, and unresolved outcomes. - [Before the first strangler cutover: prove PostgreSQL lock conflicts and read visibility, then move traffic](https://atomtechnologies.ai/insights/strangler-cutover-lock-and-visibility-gate-postgresql/): Approve the first strangler slice cutover only after a rehearsal that records the exact SQL statement, its documented lock mode, the commands it conflicts with, and the isolation level. This article gives the one-page ga - [Strangler migration: choose the first business path to move](https://atomtechnologies.ai/insights/strangler-migration-first-business-path/): Use an invoice-download example to choose a migration boundary. Separate rendering from invoice authority, find hidden callers, and design a coexistence rehearsal before funding the extraction. - [First strangler slice cutover: what to prove in PostgreSQL before production writes move](https://atomtechnologies.ai/insights/strangler-slice-cutover-gate-postgresql/): Do not cut over production writes after single-threaded parity tests alone. This article sets out the concurrent-write parity, isolation-level and lock evidence to collect first, who owns each step, and when the recommen - [Tail latency: separate queueing delay from service time](https://atomtechnologies.ai/insights/tail-latency-queueing-service-time/): Before tuning a slow request path, locate the time spent waiting for a worker and the time spent inside it. A worked timing example shows what the measurements can justify and why stage percentiles cannot be added. - [What a Webhook 2xx Acknowledgement Should Commit in a Ledger Sync](https://atomtechnologies.ai/insights/webhook-2xx-commit-point-ledger-sync/): For a firm syncing vendor webhooks into an accounting or practice-management ledger, the 2xx should commit only durable receipt of the event, with the ledger effect applied idempotently afterward. Crash, duplicate and co - [API idempotency: what survives a retry?](https://atomtechnologies.ai/insights/what-survives-a-retry/): How command identity and recorded outcomes make API retries safe. Inspect a Rust reference study and the persistence boundaries it does not establish. ## Optional - [Durable transaction study](https://atomtechnologies.ai/engineering/durable-transfers/): Run a Rust study with durable receipts, separate-request recovery, concurrent duplicates, and content conflicts. Inspect synthetic transfers in a bounded Cloudflare Durable Object. - [Transaction integrity reference study](https://atomtechnologies.ai/engineering/transaction-integrity/): Run an inspectable Rust transaction study: exact replays, identity conflicts, and recovery after a simulated lost acknowledgement. - [Privacy](https://atomtechnologies.ai/privacy/): How Atom Technologies handles website inquiries and business prospecting data, how long it is kept, and how to object or unsubscribe. - [Terms of service](https://atomtechnologies.ai/terms/): Terms of service for Atom Technologies, a trading name of Atom Digital Technologies LLP: services, fees and payment, delivery, refunds and cancellation, and governing law. ## For AI agents and assistants An automated system answers here, not a person; a person on the team reads every inquiry. - [Agent guide and privacy notice](https://atomtechnologies.ai/agents/): how to send an inquiry for someone, and what happens to the data. - [Inquiry API](https://atomtechnologies.ai/.well-known/agent-inquiry.json): POST https://atomtechnologies.ai/api/agent-inquiry with JSON (OpenAPI at the link). Returns an acknowledgement and the booking link. - MCP server (Streamable HTTP): https://atomtechnologies.ai/mcp. Tools: get_company, get_services, get_pricing, get_faq, request_quote, book_call, send_inquiry. - A2A agent card: https://atomtechnologies.ai/.well-known/agent-card.json - [Book a call](https://atomtechnologies.ai/book/?utm_source=ai_agent&utm_medium=agent&utm_campaign=agent_channel&utm_content=llms): 30 minutes with the Atom Technologies engineering lead. - [Full details for LLMs](https://atomtechnologies.ai/llms-full.txt): services, prices and the FAQ. Sitemap: https://atomtechnologies.ai/sitemap.xml