We’re a 6-person consultancy that caps engagements at eight at a time. That math means we hire a senior engineer once every six to twelve months — never to fill a seat, always to make the firm meaningfully better. If you’d be a fit, we’d like to know about you before we’re hiring, not after.
Updated weekly. When a role opens, it appears below with full editorial treatment — never as a one-line job ID with a checkbox grid. Until then, the talent network application is the way in: we read every one of them, file good notes, and reach out when a fit appears.
Most career pages list everything they want. We’d rather list what we won’t hire for. If any of this rules you out, we’d rather you find out now than at the offer stage.
None of these roles are open right now. We’ve written them out so the bar is visible — and so when an opening appears, the description is ready, not bolted together by HR in a hurry.
You have carried a pager for a real product. You can read a postmortem and tell what went unsaid. You write runbooks people actually use.
Lead 3–4 concurrent cloud infrastructure engagements per year — each one running 8–16 weeks. You’d be the named engineer on every proposal, the person on the first call, the person writing the Terraform, the person on the handover. No staffing layers between you and the client.
You’ve already shipped this kind of work somewhere serious. Probably in-house for a Series B+ company, possibly at a more senior level than you’d like to admit. You can write a paragraph about a system without it sounding like marketing. You’ve been on call for at least one workload that mattered.
Writing. You’ll write a weekly summary for every client every week. If “200–400 words on what shipped this week” sounds like a chore, this firm will be the wrong place. If it sounds like the obvious move, we’ll get along.
Run reliability engagements for Series B–C teams where the on-call rotation is bleeding senior engineers. SLO design with product teams in the room, alert pruning that sometimes deletes 1,000+ rules, incident review facilitation, and the unglamorous postmortem-template work that turns “we had a bad week” into “here’s the one process change we’re making.”
You’ve run SRE at a place where reliability mattered — not in the abstract sense, but in the sense where a sustained outage cost the company real money or real customers. You’ve been the senior person on call. You know that the engineering work is 30% of the job and the process / political work is the other 70%.
Empathy with on-call engineers. If your instinct is “they just need to learn to deal with it,” we’re not aligned. Most of our reliability work is making the human cost of on-call lower, not just the page-rate lower.
Lead AI infrastructure engagements for AI-native startups and Series A–C teams running into the Bedrock-cost wall. vLLM tuning, Triton routing, GPU-aware Kubernetes, fine-tuning pipelines, eval harnesses. The work that takes a team from “we just turned on Bedrock and the bill is scary” to “we run our own inference and the cost curve is linear.”
You’ve already run LLM inference at production scale somewhere — not just tutorial-level vLLM, but the version where you’ve tuned --max-num-seqs, debugged a chunked-prefill regression, and watched a router’s connection pooling get blamed for vLLM’s p99 numbers. You can write a paragraph about PagedAttention without sounding like a paper summary.
The eval-before-infrastructure discipline. Every migration we’ve done has succeeded because we built the eval harness before the vLLM cluster. If your instinct is “let’s ship the cluster first and figure out eval later,” we’ll get along worse than you’d hope.
We’re honest about the trade-offs of each, because too many companies pretend the modes are interchangeable. They aren’t. Pick the one that matches how you actually want to work.
A small office in Gulberg, Lahore. Mostly used by Hassan and Maryam day-to-day. Quiet, technical, no open-plan theater. We don’t mandate office attendance — the people who like working in offices choose this; the people who don’t work remotely.
We hire remote engineers anywhere with reasonable overlap with PKT business hours. That typically means UTC–3 through UTC+8 work. David is in Lisbon, our last hire is in Istanbul. We won’t pretend North America time zones work for this team.
Four stages. We aim for two to three weeks total. Most companies hide their hiring process from candidates — we’d rather you know exactly what you’re committing to before you commit.
You send us a few paragraphs — not a resume, prose. We read it, reply within 72 hours, usually with two or three follow-up questions. If we’re not a fit at this stage we tell you — with a real reason, not a form rejection.
45-minute video call with the lead engineer for the role. No leetcode. We’ll talk through a real production incident we hit recently and ask what you’d have done. You should leave the call knowing more about us, not less.
A two-hour technical exercise based on a redacted real client problem we’ve solved. We pay $200 for your time whether or not we move forward. You keep your work. We never reuse exercise submissions in client work.
A 30-minute conversation with the founder. Comp, start date, equity, expectations. You get an answer within 48 hours, with a written rationale either way. If it’s a no, we’ll tell you what would change our mind for next time.
Not perks. Operational reality. The kind of thing you’d want to know on day three, written down on day zero.
Six senior engineers means the org chart is mostly flat. You report to the founder for administrative things (compensation, time off, etc.) and to no one for engineering. You’ll be expected to act like a peer of the founders from day one — including pushing back when you think they’re wrong.
Not in parallel. We rotate engineers off engagements as cleanly as we rotate them on. Between engagements you’ll have a week or two of internal time — not vacation, but time for writing, tool-building, or just sharpening the saw. Nobody’s billable target is set per quarter.
The proposal that goes to the client has your name and a link to your LinkedIn. If you don’t want to be public-facing in that way, this firm isn’t the right fit. The accountability and the credibility are both real — you get the engagement, and the client knows exactly who they’re getting.
Every engineer at the same level makes the same band, within $10k. Equity participation in firm profit is the same percentage for everyone at the same level — not negotiated case-by-case. We share the firm’s revenue and cost numbers quarterly with the team, including margins per engagement.
No formal day count. The expectation is that you take at least 25 days a year off. We’ve had engineers take 6 weeks. We’ve never had a problem with anyone taking too much — the failure mode here is taking too little. The founder takes 5 weeks a year and tells everyone, on purpose.
For the engagements where on-call is part of the work (Reliability retainers, some AI Infra), the rotation is shared across the team — not stacked on the engineer who “owns” that engagement. Nobody carries the pager every week. We rotate weekly during retainer periods, paid separately from base.
Weekly client summaries. Postmortems. Proposals. Internal memos. If you don’t enjoy writing, this firm will exhaust you. If you do, you’ll find that the writing surfaces the thinking — and the thinking is most of what gets paid for here. We review writing samples seriously at interview.
Three real fields. We don’t need a resume PDF, a cover letter, or a list of references. We need to know what you’d bring, in your own words.
No open positions right now. This form puts you on the list we read first when an opening matches. Most hires we make come from this list.
The link to your work matters more than your resume. A GitHub profile we can read for an hour. A blog post you wrote. A talk you gave. Anything that shows your thinking, not just your job titles.