Home/Careers
Careers · AI, Salesforce, Full-Stack, Delivery

Build systems that carry real customers. Not demos.

Corelynx builds and runs software for US mid-market companies. Teams are small and senior-led, which means you own a problem end to end — you design it, you ship it, and you are the person the client talks to about it. If that sounds like more responsibility than you want, this is honestly not the place.

Apply now → See what we build
Direct answer · What is it like to work at Corelynx?

Corelynx is a 15-year software consultancy with offices in Dublin, California and Kolkata, India, and roughly 350 delivered systems behind it. We hire across four disciplines: AI and data science engineering, Salesforce, full-stack architecture, and technical project management. The work is client delivery rather than internal product, and teams are deliberately small — the person who designs a system usually ships it and then supports it. Most engineering roles are Kolkata-based, with client-facing leadership in the Bay Area.

Open disciplines

Which roles are we hiring for?

Four disciplines, hired continuously rather than on a schedule. If yours is here and there is no posting, apply anyway — we keep strong applications on file rather than restarting a search every quarter.

AI & DATA SCIENCE

AI and Data Science Engineers

Building agent systems that survive production: tool-calling, retrieval over messy client data, evaluation harnesses, and the guardrails that stop an agent doing something expensive. You should have opinions about when an agent is the wrong architecture — most of the time, on current evidence — and be able to defend them.

SALESFORCE

Salesforce Engineers and Consultants

Apex, LWC, Flows and Agentforce, on orgs that have accumulated a decade of undocumented decisions. The interesting work is not configuration; it is deciding what the data model should have been and getting there without breaking the business that runs on it.

ARCHITECTURE

Full-Stack Architects

Owning the decisions that are expensive to reverse — data model, integration boundaries, auth, environment strategy — across TypeScript, Python, Java, .NET and cloud-native platforms. Stack-agnostic by design, so breadth and judgment matter more than depth in one framework.

DELIVERY

Technical Project Managers

Running an engagement end to end: scope, milestones, client communication, and the adoption work that decides whether a system is actually used. Technical enough to challenge an estimate and to know when a developer is describing a real constraint rather than a preference.

What we look for

What actually gets you hired here?

We are not scanning for keywords. These are the things that decide it, in roughly this order.

  • Evidence you have shipped something real. A link beats a CV bullet. Production means it served users who were not you.
  • Judgment under ambiguity. Client work arrives underspecified. We want the trade-off you chose and why, not a textbook answer.
  • Depth on AI agents, or genuine curiosity about them. Not prompt-engineering — tool-calling, retrieval quality, evaluation, and knowing where the failure modes are.
  • You can explain a technical decision to a non-technical client. Most engineers here talk to clients directly. It is faster than an account layer and it is not optional.
  • Self-direction. Small teams mean nobody is assigning you tickets. The people who do well here decide what matters next and say so.
The honest version

Who is this genuinely not right for?

Published for the same reason our client pages publish where our services are the wrong answer: it saves everyone a month.

  • You want to work on one product for years. This is consulting — several clients, several stacks, and systems you hand over.
  • You want to specialise deeply in a single framework. Stack-agnostic delivery means breadth. Depth is valuable but it will not be the whole job.
  • You prefer clearly-scoped tickets. Ambiguity arrives with the work here. That is the job, not a symptom of poor planning.
  • You would rather not talk to clients. Most engineers do, regularly.
  • You want AI work that stops at the demo. Anything we build gets deployed, monitored and supported. The interesting problems start after the demo.
How we hire

What does the process look like?

Four steps, usually two to three weeks end to end. You will know where you stand after each one, including when the answer is no — silence after an interview is a choice a company makes, and we think it is a rude one.

1. A conversation with someone who does the work. Not a recruiter reading a checklist. You should leave it knowing what you would actually be building and who you would be building it with.

2. A practical exercise. A scoped problem close to real client work, in your own time, usually a few hours. No algorithm puzzles — we do not hire people to reverse binary trees, we hire them to design systems that survive contact with a client's data.

3. A technical conversation about what you built. We go through your reasoning and your trade-offs. A deliberate compromise you can defend is more interesting than a perfect answer you cannot explain.

4. Offer and expectations. Scope, team, compensation and what the first ninety days look like. If the fit is wrong we say so here rather than after you join.

The work itself

What will you actually be doing?

Client delivery, which means variety and real deadlines. In a given quarter an engineer here might design the data model for a lending CRM, build a retrieval layer over a client's document store, and rescue a Salesforce org that three previous admins left undocumented.

The AI work is the fastest-growing part and the least like the demos. Building an agent that qualifies leads inside a client's real CRM means confronting data that disagrees with itself, defining what the agent may do without a human confirming, and being able to reconstruct any decision it made six months later. That last one catches most teams out.

You will not be maintaining someone else's ticket queue. Teams are small enough that your work is visible, and unhelpful process does not accumulate because there is nobody whose job is to add it.

Apply

Apply for a role

One form, so we know exactly which role you are applying for and can compare applications fairly. The written questions are the part we actually read — take your time on them and skip the cover letter.

We reply to every application, including the ones that are a no. Recruiters and outsourcing agencies: this form is for individual applicants only.
Candidate questions

Questions candidates ask most

Most engineering and delivery roles are based in Kolkata, India, at our Salt Lake office. Client-facing and leadership roles are in Dublin, California. We say this up front because the alternative wastes your time — a US-based applicant for a Kolkata engineering role is not a fit we can make work.

The team works flexibly and delivery is remote-first for clients, but engineering roles are attached to an office rather than fully distributed. Small senior-led teams depend on a lot of unscheduled conversation, and that is harder to replicate across time zones than it is worth pretending otherwise.

Yes, if you are genuinely curious and can show us you have gone further than a tutorial. Say so honestly in the form and tell us what you would build first. We would rather have someone who admits the gap and thinks clearly than someone who overstates experience we will discover in week two.

A scoped problem resembling real client work — typically designing and partly implementing something that has to integrate with a system you do not control. A few hours, in your own time. We review the trade-offs you made rather than scoring against a rubric, and we give feedback either way.

Yes, but not in volume. Small senior-led teams mean a junior gets real mentorship and real ownership early, and it also means we can only take on as many as we can properly support. If you are early in your career, the exercise and your written answers matter far more than your CV.

Across the company: Python, TypeScript, Java, .NET, PHP, React and Node, plus SQL and data modelling; LLM APIs, RAG and agent frameworks on the AI side; Apex, LWC, Flows and Agentforce on Salesforce; and AWS, GCP or Azure underneath. On any single engagement, whichever of those fits the client's problem and their ability to hire for it afterwards.

Two to three weeks from first conversation to offer, assuming schedules line up. If it will take longer we tell you why rather than going quiet.

Pick the closest one and say so in the last written question. We would rather read one considered application than two hedged ones.

No. This form is for individual applicants. Agency submissions are not accepted and unsolicited candidate CVs do not create a fee arrangement.

Apply

Ready to apply?

One form, four disciplines. The written questions are what we read first.

Go to the application form →
First

Want to see the work?

Eight representative builds from 350+ delivered systems — bespoke CRMs, unified data platforms, agent fleets and custom applications.

View the portfolio →
Book a Strategy Session