Spell out what you need: Rigel writes the JD, scores real skills, and shows you what to sharpen next time.
Set what the role actually needs, and decide how much each part counts. Not a default someone else picked.
Publish the role and share it anywhere. Whoever opens it comes back measured on your weights.
Every score comes apart into the sections that made it — while the things that invite bias stay hidden.
Because everyone is scored the same way, the pool becomes countable — and shows you where it runs thin.
Before Rigel can match anyone, the role has to be defined right — so we made that the most honest part of hiring.
Most postings start life as an old one. A req gets reopened, last year's description gets lightly edited, and nobody re-asks what the job actually needs today.
Then the hardest translation in hiring happens off the page: a hiring manager knows what the work demands, and the person writing the posting has to turn that into requirements — often for technical work they can't evaluate themselves.
say they hire on skills — but fewer than 1 in 700 hires actually changed.
of recruiters can't tell a hiring manager's must-haves from the nice-to-haves.
of technical roles are screened by non-technical recruiters.
of jobs labelled “entry-level” actually require 3+ years.
The fix starts with writing it down — not a recycled paragraph, but explicit, structured criteria.
The title, the skills you require and the ones that are merely nice to have, the experience, the education. You set the level for each skill, so “knows Python” stops meaning four different things to four different people. That structure is what makes an honest match possible: Rigel scores against the criteria you set, not against how well someone guessed the keywords.
Decide how much skills, experience, and education count for each job — a transparent weighted sum you control, never a default someone else picked.
Every match saves the weights it used, so a score is always yours to reproduce and defend.
Skills weighted highest — a skills-first role.
Candidates scoring below this threshold won't see this job listing.
Publishing gives you two things: a job description written from the criteria you just set, and a link that turns anyone who opens it into a scored, structured candidate — measured on your weights, not on how well they formatted their CV.
Post a job anywhere else and you get a pile of PDFs — written to beat a filter, formatted forty different ways, impossible to compare, and someone has to read all of them before anyone knows anything.
Through Rigel’s link, the same person arrives as structured evidence, already scored against the criteria you set.
Results-driven and highly motivated senior software engineer with over six years of hands-on commercial experience designing, building, shipping and maintaining scalable, resilient backend services across Python, Go and TypeScript, with a proven track record of delivering high-impact, business-critical projects on time and under budget in fast-paced agile environments where requirements shifted weekly and the roadmap was rewritten twice a quarter. Led the migration of a monolithic billing platform to a distributed microservice architecture serving twelve million requests per day, reducing p99 latency by forty-one percent and cutting infrastructure spend substantially year on year while maintaining backwards compatibility for four external integrators throughout the transition. Collaborated cross-functionally with product, design and data science stakeholders to define requirements, scope deliverables and drive alignment across five engineering teams operating in three time zones. Comfortable owning ambiguity end to end, equally at home in the weeds or in front of a steering group, and a firm believer that the best documentation is the code itself.
Owned the payments ingestion pipeline.
Designed and delivered an event-driven ingestion pipeline processing roughly four terabytes of transactional data daily, introducing consumer-driven contract testing between eleven services, which reduced integration defects materially across the wider platform organisation and became the template later adopted by two adjacent groups without any formal mandate from the architecture function.
Mentored four junior engineers; two promoted inside eighteen months.
Wrote and maintained the runbooks.
Built and maintained CI/CD pipelines using GitHub Actions, Docker and Terraform, reducing median deploy time from thirty-eight minutes to under six.
Various.
Supported the on-call rota and led three blameless post-incident reviews, one of which produced a change to the deployment gate that is still in force.
Assorted internal tooling, scheduled reporting jobs, and one intranet redesign that was never shipped.
BSc (Hons) Computer Science, Second Class Honours (Upper Division), University of the North Weald, 2018. Relevant coursework: distributed systems, machine learning, compilers, database internals, discrete mathematics. A-Levels: Mathematics (A), Physics (B), Economics (B), 2015.
Python · Go · TypeScript · PostgreSQL · Redis · Kafka · FastAPI · Django · gRPC · REST · GraphQL · Docker · Kubernetes · Terraform · AWS · GCP · Prometheus · Grafana · OpenTelemetry · CI/CD · TDD · Agile · Scrum · Kanban · stakeholder management · technical writing · mentoring · code review · pair programming · public speaking · Excel · Jira · Confluence · Figma (basic) · Git · Bash · Linux · nginx · RabbitMQ · Celery · Airflow · dbt · Snowflake · Looker
No description yet.
Generate a candidate-facing description with AI to share alongside your apply link.
Generate with AIYou already told Rigel what the role needs. An LLM turns that into a clean, LinkedIn-ready description — one you can edit, regenerate, or copy out.
It writes the prose; it doesn't invent the requirements, and it has nothing to do with how anyone is scored. The matching ran on your criteria. This is just the version people read — the shortlisted candidate, when the role finally unlocks for them, and anyone else you send it to.
Every published job comes with its own apply link — paste it into the “Apply” field when you post the role on LinkedIn, on your careers page, or straight into a message.
Whoever opens it uploads a CV once and is scored against exactly the same criteria and the same weights as everyone else, so you're never comparing two different measurements. They don't need an account to apply. And if they sign up later, the profile they already gave you carries over instead of starting again.
Paste the apply link into the "Apply" field when you post the job on LinkedIn or your careers page.
And it compounds. Once someone is in Rigel, every job you publish after this scores them automatically — you never source the same person twice.
Most hiring tools hand you a number and expect you to trust it. Ask why someone ranked fourth and you get a shrug — or a paragraph of prose standing in for a number you can check.
Open a job and you get a list of people in score order. No names, no photos, no contact details — every candidate is a handle and a set of measurements, because that is all the scoring ever saw.
You are ranking evidence, not a person you have already formed an opinion about.
Click any row and the drawer shows how the number was built: each section score, the weight you set for this role, and the two multiplied out.
Every score is stored with the weights and the scoring version that produced it, so a match from three months ago still reconciles today. The arithmetic is written out, and you can reproduce it on paper.
Applied ML engineer with four years building and shipping production model services. Owns training pipelines end to end and has taken two LLM features from prototype to production. Works close to the serving layer — evaluation, latency and rollback — rather than in notebooks.
Owns the model-serving stack: training pipelines, evaluation harness and rollout. Took two LLM features from prototype to production, including the retrieval layer behind the internal search tool.
Built forecasting models for demand planning and the reporting pipeline that fed them.
Maintains a small library for regression-testing prompt changes against a fixed eval set.
An internal write-up setting p95 targets for online inference, adopted as the team’s rollout checklist.
Spoke at a regional Python meetup on regression-testing prompt changes against a fixed eval set.
Four of six required skills are evidenced at or above the level the role asks for, with Python and PyTorch supported by shipped production work rather than a skills list. Kubernetes and distributed training are not mentioned anywhere in the CV.
Four years against a target of six. All four are in applied ML rather than adjacent data roles, so the shortfall is duration, not domain.
BSc Computer Science, 2021. The role asks for a postgraduate qualification in a quantitative field; the degree is adjacent but doesn't meet the stated requirement.
Two tabs on the same candidate. The profile is the evidence — the skills and the levels they’re held at, the experience, the projects — opened all the way up, and still no name, no photo, no contact details attached to any of it. The explanation says which of it counted.
Which required skills matched and at what level, which are missing, and how far the experience sits from what you asked for — not a similarity percentage standing in for a reason. It describes what it found. It does not tell you who to hire.
That last point is where the bias reduction actually comes from. Blind screening doesn’t erase bias — a CV still carries signals. It removes the ones that should never have been in the decision in the first place.
Post a role anywhere else and you learn nothing from it. The applicants aren’t comparable, so nothing aggregates — and the next req starts from the same blank page as the last one.
Because everyone who reaches this job is scored on the same criteria and the same weights, the applicants stop being a stack of documents and become a set of measurements. That’s the thing a pile of PDFs can never be: countable.
So after the job goes live you can see the shape of who it attracted — not just who ranked highest, but what the whole pool was strong and weak at, and whether the requirements you wrote were the ones that mattered.
Which is the part hiring normally skips. The role is where hiring breaks — and this is the only place that tells you how to write the next one better.
From defining what matters to ranked candidates to the score behind each one: the whole flow, start to finish.
Post a job. See ranked candidates with full, explainable scores.
Hire by skill →