Hiring
AI Engineer — production agents, RAG and observability
Hiring evidence involving Python, FastAPI, PostgreSQL. First observed 5 Sept 2026; the response opens on upwork.com.
- Budget
- $3,000 fixed-price
- Work format
- Hiring
- Role family
- AI / ML engineer
- Seniority
- Level not stated
- Observed
- 5 Sept 202612 days old
- Response destination
- upwork.com
Source job details
Structured facts are published only when supplied by the source; missing values are never invented.
- Hiring organization
- BIX Tecnologia
- Source reference
- OPP-20260905-719
- Location evidence
- Brazil / South American time zones
- Employment type
- full time
- Required skills
- Python, Django, FastAPI, REST APIs, PostgreSQL, message queues, async processing, LangChain, LangGraph, RAG, embeddings, pgvector/vector databases, OpenAI, Claude, AWS Bedrock, Docker, CI/CD, evals, observability, guardrails
- Expected duration
- Ongoing project and contract-to-hire; hours and initial milestone duration not published
What this signal says
This is a delayed public reading of the demand behind “AI Engineer — production agents, RAG and observability.” It is deliberately smaller than the private workspace record and deliberately more interpretive than the source listing. The page identifies the work, translates its technology language into Leadiy’s controlled catalog, names the commercial format and preserves the original application destination. It does not publish personal contacts, internal qualification, prepared outreach or hidden source fields. Read it as an evidence note about one observed opportunity, with the source link retained for the current authoritative details.
How the technology was normalized
The controlled technology reading for this record is Python, FastAPI, PostgreSQL, pgvector, Docker, AWS, OpenAI API, RAG, Monitoring and alerting, Python/FastAPI + OpenAI + pgvector. That label is derived from explicit terms in the role, primary technology and stack fields; it is not inferred from the buyer’s identity or from private contact context. Normalization folds common spelling and naming variants into one catalog so this page can be compared with other records. A combination tag appears only when every required part of that combination is detected. The result is a search aid, not a guarantee that every named tool is equally central to the delivery scope.
Commercial format
Leadiy reads the buying motion as Hiring. This matters because the response shape for a deliverable, contractor seat, formal proposal and employment application is not interchangeable. The category is a classifier output rather than a legal characterization, and the original post wins if its terms change. The available duration wording is “Ongoing project and contract-to-hire; hours and initial milestone duration not published.” Work-location context is recorded as “Upwork labels the role worldwide, but the brief explicitly requires the candidate to reside in South America; agency eligibility is not stated.”
Reading the budget
The financial signal is “$3,000 fixed-price,” exactly at the level available in the normalized record. No midpoint is invented and no missing value is estimated. This restraint is important when similar technology work can be sold as fixed delivery, hourly capacity, a retainer or employment. The number is useful for triage only after its unit and scope are understood. The application destination should be used to verify what the amount includes, whether it is negotiable, and whether platform or payment conditions change the supplier’s effective return.
Demand context from Hiring Pulse
In the current 30-day Hiring Pulse, Python appears in 338 of 1254 observed opportunities (27%). Counts are multi-label, so one opportunity may contribute to several technology buckets. Hiring Pulse is the contextual layer Leadiy adds beyond aggregation. It places the record against a verified observation window while keeping counting rules visible: technologies can overlap, the feed is not a census, and sparse groups are not disclosed as exact numbers. This evidence can support prioritization, not a forecast. The record still needs source-level verification, and the aggregate view deliberately omits direct leads, personal information, private outreach content and arbitrary micro-segments that could reveal too much.
Observation and publication timing
Leadiy first observed this signal on 5 Sept 2026, 09:28 UTC; at page render it is 12 days old. Its public eligibility time was calculated once from that observation using a deterministic 46.94-hour delay inside the promised 24–48 hour window. The delay is not the source publication age and should not be presented as one. It creates a clear product boundary: authenticated customers can work from the live matched feed, while search engines and unauthenticated visitors receive a later evidence edition. The stored eligibility timestamp prevents an early publish after restarts or repeated syncs.
What looks similar
The related-record layer searches already published pages for overlap in the controlled technology catalog. It does not inspect private contacts and it does not use a buyer name as a recommendation signal. The closest current titles are: "AI Full-Stack Developer — RAG-based assistant platform", "RAG chatbot for Fine Architects", "Full-stack Next.js developer for RAG chatbot and API integrations" Treat them as adjacent demand signals, not substitutes for reading the original post. Their engagement format, budget unit, eligibility and delivery expectations may differ even when a framework or platform matches. This page keeps that uncertainty visible rather than manufacturing a single “typical” opportunity from unlike records.
Where the response goes
This page’s action route is explicit: it opens Upwork on upwork.com. Leadiy provides context and provenance, not an application proxy. A click can therefore lead to a third-party login or proposal workflow governed by that service’s own privacy, eligibility and payment rules. The public catalog never replaces those instructions and never publishes a hidden email address as an alternate route. Review the destination hostname, confirm that the opportunity is still active, and respond only through a path you trust.
Public evidence versus earlier access
The commercial distinction is one sentence: the same opportunities, one day earlier. The paid workspace is for acting on newly matched signals with private workflow tools; this public edition is for later discovery, citation and market context. Paying does not improve the source’s terms or guarantee a response. It changes when the Leadiy record becomes available and adds the operating context needed to triage it. That boundary keeps the open catalog genuinely useful while preserving a simple reason to subscribe without hiding where the original response must be submitted.
The same opportunities, one day earlier.
See the $39 / 30-day plan →Your response leaves Leadiy.
The button opens upwork.com. That third-party page controls login, eligibility, proposal fields, privacy and the current status of the opportunity.