Your first ten engineers will write the first version of everything: the codebase, the on call habits, the way the team argues about trade-offs. You are also hiring them with no recruiter, no applicant tracking system and a calendar full of customer calls. Candidates are not the scarce resource. Your time is, along with your ability to tell a strong engineer from a convincing one.
This playbook is a seven step screening process built around that constraint. It runs on a spreadsheet and a few calendar invites. It draws on published advice from Y Combinator and eChai Ventures, one meta analysis of which selection methods predict job performance, and the realities of hiring in India, from notice periods to equity. Where a source has a commercial interest, the text says so. Sources were checked in October 2026.
Why the first ten hires are different
In a large company a weak hire is a statistic. In a team of five it is a quarter of your engineering capacity. eChai Ventures says as much in its guide for first time hirers: a wrong early hire is “disproportionately damaging” in a tiny team.
That has practical consequences. You have too few data points to calibrate gut feel, so you need a written scorecard. Every stage of a long funnel costs founder time, so the funnel has to be short. And because you will see fewer candidates than a scale up does, each stage has to produce a real signal rather than act as a volume filter.
Step 1: Write the scorecard before the job post
Before you contact anyone, write down five or six things the person must be good at, and what “good” looks like. Harj Taggar’s Y Combinator guide (2018) makes the same point: list the specific criteria your ideal hire would meet, cover technical ability and how well you would work together, and score each candidate against them. Taggar was then CEO of Triplebyte, a hiring marketplace, which the article discloses.
Here is an example for a first backend engineer. Treat it as a starting point and replace the criteria with your own:
| Criterion | What good looks like | Where you test it |
|---|---|---|
| Core stack depth | Explains why they chose past designs and what broke | Work sample, structured interview |
| Ownership | Has shipped something end to end and fixed it in production | Founder call, references |
| Product judgment | Asks what problem a feature solves before building it | Founder call, work sample |
| Communication | Writes and explains clearly, asks sharp questions | Work sample write up |
| Learning speed | Picked up an unfamiliar tool or domain quickly | Structured interview |
| Reliability | Does what they said, when they said it | Process itself, references |
Give each criterion a score from 1 to 4 with a one line anchor, for example 1 for “could not explain it” and 4 for “taught me something”. Write the anchors now, while you are not looking at a charming candidate. Do not change the criteria halfway through a search unless you record why.
Step 2: Start with your network, then widen
The Y Combinator guide calls personal networks the most important source by far and warns that working them takes time, at least a third of yours. List the best engineers you know, ask whether they are interested, then ask who they would hire. eChai’s roundup on hiring in India points the same way. It recommends alumni, ex colleagues and referrals first, job portals as one channel among several, and a referral bounty to turn your team’s contacts into a sourcing channel. It also advises selling the mission and the problem rather than the title or the salary.
A network has a cost. It tends to return people who look like you and your co-founders. Keep at least one open channel, such as a public job post or a developer community, so you see candidates outside your circle. Whichever channel a candidate comes from, they go through the same scorecard.
Step 3: Check resumes against three must haves
Pick three must haves from your scorecard and read each resume for those only, spending five minutes or less on each. The same eChai roundup, which draws on other publishers, advises weighting scrappiness over a marquee employer when hiring for early stage Indian startups. Prior startup experience and signals such as side projects, leadership and hard problems tackled may tell you more than a brand name employer or a degree. Take that as one experienced view rather than a rule, but it is a good reason not to filter on college or company name alone.
Write one line on why you advanced or declined each person. You will need it to spot patterns later, and it keeps the decision tied to your criteria.
Step 4: Run a 20 minute founder call
Use the same five or six questions for every candidate, in the same order, and score as you go. Ask about specific things they built, what went wrong and what they would do differently. Save the pitch for later, once you have decided you want the person. A call where you sell for fifteen minutes tells you nothing about the candidate.
Step 5: Test with real work
eChai’s guide for first time hirers recommends a paid, real world work sample that mirrors the job, in place of trivia or whiteboard tests. The research is consistent with this. In their 2022 meta analysis in the Journal of Applied Psychology, Sackett and colleagues estimated operational validities of .42 for structured interviews, .40 for job knowledge tests and .33 for work sample tests, as summarised by SIOP. These are averages across many jobs, not promises for any one hire. The sensible reading is that no single stage is enough, and that a job like task combined with a structured conversation beats either alone.
Practical rules for the work sample:
- Scope it to two or three hours and pay for the time. Pay in rupees against a simple invoice or stipend, and check the paperwork with your accountant.
- Use a slice of your real problem, such as a small API with one awkward requirement, not a puzzle.
- State your AI policy in the brief. Candidates will use AI tools unless you say otherwise. Allowing them is a legitimate choice, but then judge the reasoning, not the typing.
- Follow it with a 30 minute walkthrough in which the candidate explains their choices and changes one thing live. It is the strongest check that the work is theirs.
Step 6: Hold one structured interview
For one or two engineers on your team, or you plus a trusted advisor, run a single 45 to 60 minute interview built from the scorecard. Every candidate gets the same core questions, and each interviewer scores independently before discussing. The aim is the structured format that Sackett’s estimates favour: consistent questions, anchored scores, and no debate until the numbers are written down.
If you are not an engineer yourself, bring in a technical advisor for this step and for the work sample review. Their job is to judge depth. Yours is to judge fit and ownership.
Step 7: Check references, then close with the notice period in mind
Always call references, and ask past managers the blunt question eChai suggests: would you hire this person again? Two calls of fifteen minutes each are enough.
Then plan around the notice period, which shapes your real timeline in India. A May 2026 Instahyre guide describes typical notice periods of 15 to 30 days at startups, 30 to 60 days at mid size product companies and 60 to 90 days at large product companies and unicorns, with IT services firms around 90 days and senior roles sometimes longer. The guide cites no dataset, so treat the ranges as indicative and ask each candidate for their contract terms.
Three things help. Ask whether the candidate’s current employer allows early release or a buyout, and offer to reimburse the buyout if you can. Stay in touch through the notice period with an onboarding plan, a first project and a team call. And explain your equity clearly: share count, percentage of the company, current valuation, vesting and tax treatment. For context, the eChai roundup cites a Qapita guide saying early stage Indian startups typically set aside 10 to 15 percent for an ESOP pool, with four year vesting and a one year cliff. That is a common pattern, not a requirement.
The whole funnel on one page
| Stage | What it answers | Founder time (planning assumption) | Decision rule |
|---|---|---|---|
| Resume check | Does this person meet the three must haves? | 5 minutes | Advance on all three |
| Founder call | Ownership, judgment, interest | 20 minutes | Average 3 or higher on the scorecard |
| Work sample + walkthrough | Can they do the job? | 45 minutes to review and discuss | Meets the bar on core stack and communication |
| Structured interview | Depth and learning speed | 60 minutes | Independent scores agree within one point |
| References | Would others hire them again? | 30 minutes | No unexplained concern |
| Offer | Fit, notice period, equity | Variable | Written offer within days of the final decision |
The times are planning assumptions to adjust, not benchmarks. They show where founder hours go and which stage to cut first if you run out.
Keep the records tidy
A spreadsheet is enough for the first few hires: one row per candidate, the scorecard values, the reason you advanced or declined them and the date. Keep candidate data out of personal inboxes where you can, and tell candidates how you will use their information and how long you will keep it. India’s DPDP Rules were notified in November 2025, and under the phased schedule the main duties on notice, consent and security start around 13 May 2027. Check the current dates and take legal advice for your situation.
Common mistakes
- Selling before screening. You will talk yourself into someone because you liked the call.
- Changing the criteria mid search. If you move the bar for one candidate, you cannot compare anyone.
- Long unpaid take home. Good candidates with options will decline them.
- Skipping references because the person is impressive.
- Hiring only from your own network. It is fast and often good, but it narrows who you see.
- Going silent. A candidate on a 60 day notice period needs to hear from you.
When a tool starts to help
A spreadsheet stops working when screening takes a real share of your week, when more than one role is open, or when you cannot remember what you decided about a candidate last month. At that point one system that holds the pipeline, the assessments and the notes saves time. Kestrel is an AI native hiring platform for engineering teams at startups and scale ups, and it is aimed at founding CTOs and heads of people at that stage.
In practice that means an applicant tracking system with a configurable pipeline, and resume screening that scores each application against your job description and gives written reasoning. It also means multiple choice and coding rounds, and AI interviews that produce a scorecard with a recommended outcome (Advance, Hold or Reject).
Resume screening acts on its score by default. It moves candidates at or above a threshold you set (default 40) to the next round and rejects those below it, so set the threshold deliberately and read what it rejects before you rely on it. The score includes a small education component, at most 10 of 100 points, so check it is not driving your rejections. For AI interviews, the AI recommends and a recruiter or hiring manager decides. You can always book a demo.
Frequently asked questions
Should I interview every one of the first ten engineers myself?
Yes, at least for the founder call and the final decision. Early hires set the standard for everyone who follows, and the Y Combinator guide expects founders to spend at least a third of their time on network sourcing in the early days.
Is it worth using a recruiter or agency?
Sometimes, for reach. The Y Combinator guide ranks agencies last as a source of first engineers, but its author ran a competing marketplace and wrote in 2018, so weigh it accordingly. If you use an agency, give it your scorecard and keep the same screening steps.
Do I have to pay for the work sample?
It is strongly recommended. eChai specifically advises a paid, real world work sample, and long unpaid tasks are a common reason strong candidates drop out.
What if I am not technical?
Bring in a technical advisor for the work sample review and the structured interview. Use the scorecard so their view and yours are recorded separately.