Take the grind out of
the hiring pipeline.
Let Talenry source matching candidates, score every applicant against your rubric, run the first screen and fill your calendar with interviews.
See it in action.
Senior Backend Engineer
2,140 in market
- 18found
- 14new to you
- 9matched
- 6outreach drafted
Must have
PythonFastAPIPostgreSQLAsyncIO LangGraphKafkapgvector
- Elena Kovács Senior backend engineer, 8 years of async Python at scale GitHubPeople DBEnrichment 94fit
- Tomás Ribeiro Backend engineer, 6 years on high-throughput FastAPI services People DBInbound 88fit
- Lukas Berg Senior backend engineer, deep Postgres and distributed systems GitHubPeople DB 81fit
- Priya Nair Python backend developer, 5 years across Django and FastAPI People DB 72fit
- Brandon Wells Full-stack developer, 2 years, primarily Node.js Inbound 41fit
Duplicates collapsed by shared identity, not fuzzy name match. Your own talent pool was recalled first, and those recalls are free.
5 scored
Elena Kovács
Dimensions
Why
Exceptional match. Production LangGraph, FastAPI and Postgres at scale directly mirrors our stack. EU-remote and exceeds the five-year bar by three years.
Strengths
- Production LangGraph orchestration
- 8y async Python at scale
- Timezone fit
Gaps
- No Azure, which we use lightly
History
Staff Backend Engineer — Stripe Mar 2021 — present · 4y 11m Led the async payments-events service. Built an agent-orchestration layer on LangGraph.
Senior Backend Engineer — Revolut Jun 2017 — Feb 2021 · 3y 8m Owned the ledger service; scaled Postgres with pgvector-backed search.
Tomás Ribeiro
Dimensions
Why
Strong FastAPI, Postgres and Azure match with solid EU-remote fit. Slightly light on agent-orchestration experience but the fundamentals are excellent.
Strengths
- FastAPI and Azure depth
- Event-driven systems
- EU timezone
Gaps
- No LangGraph experience
- One year under the senior ideal
History
Senior Backend Engineer — Feedzai Jan 2020 — present · 6y 1m Built fraud-scoring APIs in FastAPI across EU regions.
Backend Engineer — Talkdesk Aug 2018 — Dec 2019 · 1y 4m Async call-event ingestion pipeline in Python.
Lukas Berg
Dimensions
Why
Excellent senior profile — Kafka, Postgres and FastAPI all present. In final offer stage; strong all-round backend match in-zone.
Strengths
- Distributed systems at scale
- Kafka and Postgres depth
- EU-based
Gaps
- LangGraph would need ramp-up
History
Senior Backend Engineer — Celonis Feb 2019 — present · 7y Led process-mining ingestion; FastAPI, Kafka and Postgres at petabyte scale.
Backend Engineer — Personio Jul 2016 — Jan 2019 · 2y 6m HR-tech APIs in Python.
Priya Nair
Dimensions
Why
Solid core skills but the EU-timezone requirement is a real obstacle. Borderline — worth a conversation about working hours before deciding.
Strengths
- Strong FastAPI fundamentals
- Clean API design
Gaps
- Timezone mismatch for an EU role
- No agent orchestration work
History
Backend Engineer — Freshworks Apr 2019 — present · 6y 10m Owned billing and notification microservices in FastAPI and Postgres.
Brandon Wells
Dimensions
Why
Below the five-year senior bar at two years, primarily Node rather than Python and FastAPI, and outside the required timezone. Archived with the reason recorded.
Strengths
- Ships quickly
- Comfortable across the stack
Gaps
- Two years against a five-year bar
- Not a Python-first background
- Timezone mismatch
History
Junior Software Engineer — Indeed Aug 2023 — present · 2y 6m Built internal dashboards in Node and Express.
Voice screen · Tomás Ribeiro · 11m 04s
Gaps detected before the call
- employment gap~7 months gap before starting Backend Engineer at Talkdesk.probed
- missing skillNo LangGraph or agent-orchestration experience on the profile.probed
Transcript
Talenry There is a gap of about seven months before you started at Talkdesk. Can you tell me what you were doing then?
Tomás Parental leave. My daughter was born that March and I took the full entitlement, then spent the last two months of it doing a Postgres performance course before I started interviewing again.
Talenry Thanks, that is helpful. The role leans on agent orchestration — LangGraph specifically. What is your experience there?
Tomás Honestly, none in production. I have read the docs and built a toy router, but I would be learning it on the job. What I do have is the event-driven side, which is most of the hard part underneath it.
Talenry Fair. Tell me about the uptime number on the fraud-scoring APIs — what did you personally change to hold it?
Tomás Two things. Regional failover that did not depend on a shared Postgres primary, and a circuit breaker on the enrichment calls so a slow third party degraded scoring instead of taking it down.
78/100
Recommendation: advance
Advance. The employment gap resolved cleanly — a parental leave he volunteered before being asked. The orchestration gap is real but narrow, and everything underneath it is strong.
Always on
- Protected traits are never inferred, and self-reported EEO data never reaches the scorer.
- Adverse-impact testing runs across the pipeline, not per candidate.
- Every dimension carries the sentence that justifies it, so a score can be argued with.
Interview · Tomás Ribeiro
Times are offered in the candidate timezone and written back to your ATS as the system of record.
Six things the walkthrough
does not have a screen for.
-
Your ATS stays the system of record
Talenry mirrors into the stack your team already runs rather than annexing it. The pipeline your recruiters read stays the one they read, and everything the agent does is written back.
WorkdayGreenhouseLeverAshbyiCIMS
-
Adverse-impact testing across the whole pipeline
EEOC four-fifths testing runs with disclosure-rate context. Self-reported EEO data is segregated from scoring — never fed to an agent, never shown per candidate — and protected traits are never inferred, for any reason.
EEOC 4/5thsEEO segregated
-
The people you already know, checked first
Every search recalls your own talent pool before anyone pays for an external query. Those recalls are free, they run on every search, and they surface the candidate your team already met eighteen months ago.
-
It opens the repository and reads the source
For engineering roles Talenry pulls actual source files from a candidate’s top repositories and scores them across six dimensions — readability, structure, testing, documentation, error handling and idiomaticity. Not a commit count.
-
Configurable, pausable, and it stops when told
Approval gates are a policy you set per requisition rather than an architecture you fight. Autonomy is dialled up or down per role, and the kill-switch is honored at every step of the pipeline.
-
Two of your teams never message the same person
A cross-team guard blocks a second recruiter from cold-messaging someone already in conversation. Opt-out is detected deterministically and closes the thread before any model is called, and per-requisition caps bound the spend on any single search.
Per seat, per month.
Overage bills, it never blocks.
Enterprise adds custom model tuning, an SLA, and on-prem or private cloud.
What procurement asks
before security does.
Does this replace our ATS?
No. Your ATS stays the system of record. Talenry mirrors into it rather than annexing it, so the pipeline your team already reads stays the one they read.
How much autonomy does the agent actually have?
As much as you configure, per requisition. Approval gates are a policy you set rather than an architecture you fight, autonomy is pausable, and the kill-switch is honored at every step.
How do you handle adverse-impact and EEO compliance?
EEOC four-fifths adverse-impact testing runs across your pipeline with disclosure-rate context. Self-reported EEO data is segregated from scoring — never fed to an agent, never shown per candidate — and protected traits are never inferred.
What stops two of our teams contacting the same person?
A cross-team guard. It blocks a second team from cold-messaging someone already in conversation, and opt-out is detected deterministically and closes the conversation before any model is called.
Can the scoring rubric match how we actually hire?
Yes. Scoring runs against your rubric, and you change it by describing the change in plain language rather than editing weights. Per-requisition enrichment caps bound the spend on any single search.