⚠ Names, cities and identifying details are changed (Russian data-protection law); figures are given as ranges. All cases are published with client consent; any resemblance is coincidental.
AI / ML product3D / GPU low-levelRequirement reframing
Case 1: when “CUDA required” was shrinking the funnel
The business challenge
A product ML/AI company had spent months unable to fill a rare hybrid role — 3D graphics plus low-level GPU development. The requirement read as “CUDA required” and sharply narrowed the funnel of strong candidates.
How we worked
On concrete CVs we showed the client that the task is solved by a broader competency — low-level GPU development (graphics and/or compute) — and that CUDA logically belongs in the “nice-to-have” column. We reframed the requirement into a workable form and widened the funnel. Separately, we helped close communication with a parallel internal candidate respectfully, without burning bridges.
Outcome
A strong finalist joined; after the first month the client gave positive feedback. The search stopped stalling on a marker technology.
Principle: First clarify what task a requirement serves — and only then search by technology.
ML / Data ScienceRetentionHard conversations
Case 2: reworking compensation — and keeping the specialist
The business challenge
After the acute hiring phase the project stabilised and the team filled out. A delicate request appeared — to rework the compensation of a strong but “expensive” specialist without breaking trust or triggering an exit.
How we worked
We helped turn a risky “cut the pay” request into a correct conversation: recognition of contribution, the reasons for the change, an honest assessment with growth areas, a development plan and a review horizon. We brought the dialogue back into an “assessment plus development” frame, suggested a tone without devaluation, and afterwards collected the employee’s feedback.
Outcome
The employee stayed and accepted the new terms, with clear directions for growth. A risky moment became a managed development plan.
Principle: Hard conversations hold up when there is respect, clear reasons and a plan for growth.
Project managementOffer stageEthical communication
Case 3: managing uncertainty at the offer stage
The business challenge
At the final stage a candidate (a project manager) asked for a salary review, while inside the company an abrupt withdrawal of the offer was being discussed — a step with high reputational risk.
How we worked
We restrained the impulse to “pull the offer without reasons”, explaining the consequences; we proposed gathering facts first and building a correct communication scenario. We kept the negotiations transparent, clarified what was critical for the candidate, gave the client time for a considered decision, and helped clarify the role boundaries.
Outcome
The candidate confirmed readiness to accept; the final stage passed without reputational loss or burned bridges on the market.
Principle: An offer is an instrument of trust; at the final stage facts and careful communication decide, not abrupt moves.
C++ / QtProbationOnboarding
Case 4: doubts during probation — facts instead of feelings
The business challenge
Towards the end of probation the client developed doubts about the developer’s autonomy. The risk was either cutting on emotion or dragging the decision out with an unclear outcome.
How we worked
We moved the question from feelings to facts: gathered structured feedback from the team lead, named concrete growth areas, and agreed an improvement plan with checkpoints.
Outcome
The decision became predictable for both sides — without threats or abrupt moves: the employee had a clear plan, the client had criteria.
Principle: Probation doubts are better resolved through facts and an improvement plan — then the decision is predictable for everyone.
Product analytics (junior)Screening designAnti-cheat
Case 5: screening a junior in the age of AI
The business challenge
We needed to hire a junior analyst and assess skills honestly in a setting where test answers are easily “boosted” with an LLM.
How we worked
We built a calibrated junior screen: questions with clear markers (green and red signals), video-recorded answers, and explicit management of the “answers via AI” risk. We checked systematic thinking and communication, not memorised phrasing.
Outcome
The client got results comparable across candidates and protected against gaming — the hiring decision became well-grounded.
Principle: For a junior, what matters is testing systematic thinking and communication through cases with clear markers.
3D / PythonTransparent assessmentReproducibility
Case 6: an honest, reproducible technical assessment
The business challenge
We needed to assess 3D/Python candidates so the result did not depend on “boosted” answers and was reproducible for the client.
How we worked
We built a transparent scheme: clear rules, an independent expert review, and a final validation. We did not complicate tasks for the sake of it; we made the check repeatable.
Outcome
The assessment became protected and comparable; the client decided on transparent, reproducible data.
Principle: The winner is not the one who makes tasks harder, but the one who makes the check transparent and reproducible.
Infrastructure / SysadminOffer stageOne clear line
Case 7: one clear line of communication at the offer stage
The business challenge
At the offer-and-rejection stage the client’s position became inconsistent and “wavering” — a risk of confusing the candidate and harming the company’s reputation.
How we worked
We held a single line of communication: did not “think for” the client, carefully synced the status, and closed the situation with the candidate calmly and correctly.
Outcome
The candidate received clear, respectful communication; the situation closed without negativity or reputational loss for the employer.
Principle: At the offer and the rejection, what matters most is holding one honest line — and not thinking on the client’s behalf.
More cases available on request — bound by client NDAs. Email
[email protected] to discuss your search context.
