Case
A Blockchain Engineer for a decentralized AI network in one month
We found a core protocol developer in one month in a closed market
A decentralized AI compute project and L1 protocol from the US ($60M+ invested in the protocol) needed a Blockchain Engineer to develop the protocol core, scale the network under high load and own the security of on-chain/off-chain interactions. It took one month from the start of the search to the offer, and the offer went to a candidate from the first shortlist.
Request and constraints
Our client is building a decentralized network in which participants from all over the world pool their GPUs for AI workloads. The protocol is growing fast in two directions: new compute providers are joining the network and AI inference volumes are rising, and so is the number of developers who use this compute.
The client needed a developer who writes the blockchain itself — the code the whole network runs on. A specialist who builds services and applications on top of an existing blockchain was not a fit.
The new engineer was expected to:
- Scale the protocol
- Ship network upgrades
- Look for vulnerabilities
- Work on the open-source protocol
- Work with the open-source community and with external and internal teams
Why it was hard
A rare candidate profile
The client needed an engineer who writes the protocol core itself, not applications on top of it, and who had experience in target L1 or L2 protocols. Given the client's other requirements, there are literally a few dozen such people on the market.
A closed field
LinkedIn did not work as a search tool: only about half of blockchain engineers have profiles there, and those profiles usually say nothing about which level of the protocol they work at.
A cold market
Employers hold on to strong engineers from large protocols, and the engineers themselves are often reluctant to change companies. On top of that, the target protocols the client wanted people from are bigger and more stable, and engineers from such companies are usually in no hurry to swap their job for a startup that may not take off.
How we searched
We searched by code, not by profiles
The blockchain protocols the client was interested in are open-source projects, so GitHub became the main source. Margarita drew up a list of target protocols: projects at the intersection of blockchain and AI, as well as L1 and L2 networks with sharding. She then picked the repositories that belong to the protocol core and studied who contributes to them.
You can get onto a contributor list after a single change. AI helped Margarita single out those who actually write relevant code in the protocol core: she used it to analyze commits.
We pieced contacts together bit by bit
Finding a person in a repository is only half the job. On GitHub many people list only a username, with no company and no contacts. And on LinkedIn a single protocol may be represented by several company pages at once. So each candidate had to be cross-checked across different sources to make sure it was the right person.
“It was practically detective work. You find a person in a protocol's repository, and all they have is a name or just a nickname. You look for them on LinkedIn — there's no profile, or it just says Software Engineer, and you can't tell anything.”
Margarita Askarova, IT Recruiter NEWHR
After skills and contacts were verified, about 100 people entered the funnel.
We tailored the pitch to each candidate
Each of Margarita's outreach messages was built around the experience of a specific person. To some she wrote about the technology, to others about growth opportunities or the variety of tasks: scaling, network upgrades, security.
“There are very few people who fit, and I didn't want to lose the chance to get them interested. So I studied each candidate's profile in detail and looked for what might hook that particular person.”
Margarita Askarova, IT Recruiter NEWHR
Technical challenges worked best: building a system almost from scratch, solving security problems, working with an unfamiliar language.
In the end, 34 people replied, and 11 of them passed screening with NEWHR. The role went to a candidate from the very first shortlist Margarita sent to the client.
Funnel
| Stage | Candidates |
|---|---|
| Profiles researched | ~100 |
| Replied to the message | 34 |
| Passed interviews with NEWHR | 11 |
| Presented to the client | 7 |
| Passed technical interviews | 4 |
| Received an offer | 1 |
Result
NEWHR found candidates who are not in the usual channels, presented the client with a targeted shortlist and closed the role in one month.