The problem
Finding contract work usually means repeating yourself. You rewrite the CV, fill in another application form, wait, and rarely learn why nothing came back. Capable professionals end up spending more time applying than working.
Talnify is a contract-work matching platform from Fensso Tech LLP, TechOvelpo's parent company, built by TechOvelpo engineers. A professional builds one free profile from their CV, is assessed by industry experts, and has relevant contract opportunities matched to their skills. There are no applications to fill in, and each match shows the reasons it fits.
For us this was a product-design problem as much as an engineering one. A marketplace that matches people to work is only useful if both sides trust the matching. The brief we set ourselves was: no black boxes, no public CVs, and no dead ends.
A note on what this page covers. The sources we can publish describe Talnify's behaviour and design choices, not its internal stack, team size or timeline, so we have left those out rather than guess. What follows is the product architecture and the reasoning behind it.
Constraints
- Professionals should not have to apply. The profile has to carry the weight of every application form, so it needs to be complete without being tedious to create.
- Matches must be explainable. If a person is shown an opportunity, they should see why. That requirement affects how matching is designed, not just how it is displayed.
- Privacy by default. A CV is sensitive. It must stay private until the owner chooses to share it, and sharing must be a deliberate action.
- Assessment needs a clear outcome. Experts assess skills and return a qualified or not-qualified result with feedback. A vague score would defeat the point.
- Free for professionals. Joining costs nothing, so the product has to be built around that model.
Options considered
| Decision | Options | What we chose and why |
|---|---|---|
| How profiles are created | Long manual forms, or build from an existing CV | Build from the CV. The professional already has the document, and asking them to retype it is the exact friction the product exists to remove. |
| How skill is verified | Self-declared skills, automated tests only, or expert assessment | Expert assessment with a binary outcome and written feedback. Self-declared skills are cheap but weak signals; a clear outcome gives both sides something to rely on. |
| How matches are presented | A ranked list with no explanation, or each match with its reasons | Reasons shown for every match. A person can decide in seconds whether an opportunity is worth their time. |
| CV visibility | Public by default, or private until shared | Private until the owner chooses to share. Public CVs invite scraping and put off exactly the people who are quietly between roles. |
Architecture
At the level we can describe, the platform is a pipeline from document to profile to match, with a privacy gate at the end.
CV upload
-> Professional profile (one per person, editable any time)
-> Expert assessment (qualified or not qualified, with feedback)
-> Matching against contract opportunities (reasons recorded with each match)
-> Opportunity shown to the professional
-> CV shared only when the owner chooses
| Stage | What it produces | Design consideration |
|---|---|---|
| Profile | Structured skills and experience from the CV | Must be editable so the profile stays current |
| Assessment | Outcome and feedback | Outcome is explicit, so a not-qualified result is a path to improve, not a dead end |
| Matching | Opportunities tied to skills | Each match stores why it matched, so the explanation is data, not decoration |
| Sharing | Owner-controlled release of the CV | Default state is private |
The supporting features are the CV builder, with ready-made templates, and the CV score, which runs checks and gives suggestions. They live beside the main flow and exist to improve the input the rest of the pipeline depends on.
Notable engineering decisions
Explanations are stored with the match
"Show the reasons" is easy to promise and easy to fake, by generating a plausible sentence after the fact. The honest design is to record why a match was made at the moment it is made, and display exactly that. It costs a little more in data modelling and removes any gap between what the system did and what it says it did.
Privacy as a default state, not a setting
Treating "private" as the starting condition, rather than an option buried in settings, changes how the rest of the system is written. Anything that reads a CV has to ask whether the owner has shared it. We would rather have that check be unavoidable than be remembered.
An outcome that can be acted on
A qualified or not-qualified result, with feedback, is a small product decision with a large engineering consequence: the assessment has to produce structured results that the rest of the platform can read, not just a note. It also means a not-qualified professional can return later with a stronger profile, so assessment history needs to be kept sensibly.
Tools that raise input quality
The CV builder and score look like extras. They are really quality controls on the platform's most important input. Better CVs produce better profiles, better profiles produce better matches.
What was hard, and what we would do differently
Matching quality is bounded by profile quality. No matching approach can recover information that was never captured. Our lesson is to invest in profile completeness first, through the CV import, the builder and the score, before tuning how matches are computed.
Explainability constrains the approach. If every match must come with reasons, you cannot choose a technique that cannot produce them. We would decide that constraint at the start of design rather than discover it when building the interface.
Two-sided platforms start with a cold side. A professional sees little value until opportunities exist, and opportunity providers see little value until professionals are assessed. In planning a similar marketplace, we would decide which side to serve first and keep the other side's workflow as light as possible.
Assessment is a human process inside a software product. Reviewers need tooling, queues and consistent criteria. It is easy to design the professional's screens and underbuild the expert's.
What this means if you are building something similar
A matching or marketplace product succeeds on a few early decisions: what trust each side needs, what you will never show without consent, and how you will explain every automated decision. Those are worth settling before any interface is designed.
On the build side, platforms like this are mostly ordinary web engineering: profiles, document handling, workflows, notifications and permissions. If you need people to build one, our full-stack developers cover the front end and the API, and a dedicated development team fits a product that will evolve after launch. If you want to validate an idea first with a defined scope, a fixed-price project limits risk.
Talnify is live at talnify.com. To talk about a marketplace or matching product of your own, contact us. For a different kind of platform from the same engineers, see the Krayuv procurement case study.
