# GradPipe — full content > GradPipe is an engineering hiring network that matches startup founders to engineers by the work they can prove — analyzing the whole builder record (GitHub down to the syntax tree, shipped systems, portfolio, resume, and LinkedIn) instead of credentials. > Source: https://gradpipe.com · Generated: 2026-07-21 --- # About GradPipe GradPipe is an engineering hiring network — small on purpose. We match startup founders to engineers by the work they can prove, not the credentials they can list. ## What we do Most hiring tools rank people by signals that have stopped meaning much: a résumé line, a brand-name employer, a timed coding puzzle. GradPipe ignores the proxy and reads the work itself. For every engineer on our register we review the whole builder record: - **GitHub** — public code analyzed down to the syntax tree - **Shipped systems** — what they actually built and ran in production - **Portfolio** — projects, write-ups, and technical artifacts - **Résumé and LinkedIn** — the context around the work From that we know who wrote the hard parts: who authored the CUDA kernel, who maintained the library thousands of people depend on, who designed the zero-downtime database cutover. Then, when a founder describes a role, we match against that proven work and make a small, considered set of introductions. ## Why this is different from LeetCode-style screening A timed algorithm puzzle measures how well someone performs a puzzle. In an era where AI can generate a passable solution to most of them in seconds, that signal is weaker than ever. The code an engineer has already written, designed, debugged, and shipped is much harder to fake and much closer to the job. We read that instead. ## How we work with founders - We review GitHub, portfolio, résumé, and LinkedIn for every candidate. - We return **three matches per role**, not fifty. - We reply with our **first set of candidates within a working day**. - Introductions are direct — you talk to the engineer, we get out of the way. - Most placements close in **under three weeks**. ## Who's behind it GradPipe is run by three engineers: - **Muhammad** — Co-founder & CEO. IIT Bombay, two-time founder. Runs the outbound engine that brings engineers onto the platform and decides which engineers founders meet. - **Tec** — Co-founder & CTO. Computer Science from IIT Bombay, ex-Google (pre-placement offer). Built the backend, the matching and search algorithms, and the analysis pipeline. - **Pratham** — Co-founder & CPO. MTech in AI at BharatGen, IIT Bombay. Builds the product and runs every user call. Small on purpose: every role, every register entry, and every introduction passes through one of these three. ## Get in touch - Founders hiring engineers: [find talent](#find-talent) - Engineers joining the network: [join as a builder](/candidates) - Email: contact@gradpipe.com --- # How GradPipe works GradPipe measures engineering signal by reading the work an engineer has already done, then matches that work to what a founder is building. This page explains the methodology in detail. ## The problem with credential-based hiring The conventional hiring funnel filters on proxies: which school, which previous employer, how many years, how fast someone solves a puzzle under a timer. Each proxy was once a useful shortcut. In a post-AI world they are all weaker: - A timed algorithm question now measures whether a candidate can do something a language model does in seconds. It correlates poorly with the work of designing, building, and maintaining real systems. - Brand-name employers and elite schools are noisy. Plenty of the strongest builders never went through either; plenty of weak engineers did. The signal that has held up is the work itself: code that was written, reviewed, shipped, and maintained by a real person solving a real problem. ## We read the whole builder record For every engineer on the register we review four sources together: 1. **GitHub** — public repositories, analyzed at the source level. 2. **Shipped systems** — what the engineer built and operated in production. 3. **Portfolio** — projects, technical write-ups, and artifacts. 4. **Résumé and LinkedIn** — the context that explains the work. No single source decides anything. A clean repo with no users, a big-name employer with no visible output, or a strong résumé with no shipped work each tells only part of the story. We combine them. ## AST-level code analysis The most distinctive part of our process is how we read code. Rather than count stars or commits, we parse public code into its **abstract syntax tree (AST)** — the structured representation a compiler sees — and read the engineering decisions underneath the surface text. Working at the AST level lets us look past cosmetic details (formatting, variable names a linter would fix) and at the things that actually separate strong engineers from weak ones. The categories of signal we extract include: - **Structural complexity** — how the engineer manages control flow and the inherent difficulty of what they built. - **Abstraction and architecture** — how responsibilities are decomposed, how interfaces are drawn, whether the design scales. - **Error handling and edge cases** — whether failure paths are real or decorative. - **Concurrency and performance work** — evidence of work in the hard regimes: low latency, parallelism, systems-level constraints. - **Naming and readability** — whether other engineers could maintain the code. - **Testing and verification** — how the engineer convinces themselves the code is correct. We benchmark these signals against the work of engineers who have already proven themselves at the highest level, so a score reflects how the work compares to a real bar rather than an arbitrary scale. > We disclose the categories of signal we read, not the exact weights, > thresholds, or feature definitions. Those are how the engine earns its keep. ## Authorship, not just presence Open code is full of work that wasn't done by the person who forked it. Our analysis is built to attribute the hard parts to the engineer who actually wrote them — the kernel, the protocol, the migration — rather than crediting everyone listed on a popular repository. ## How a match happens 1. A founder tells us what they're building, what's hard about it, and who they've already tried to hire. 2. We read the role against the register and reply **within a working day** with **three to five candidates** whose proven work fits — a small, considered set, not a long-list. 3. Introductions are **direct**: no applicant tracking system, no platform middleware, no scheduled nurture campaigns. The founder talks to the engineer. 4. Most placements close in **under three weeks**. ## Why this produces better outcomes When the evaluation is the engineer's real work, both sides start from substance. Founders spend their time talking to people who have demonstrably done the hard thing before; engineers are judged on what they've shipped rather than how they perform in an artificial test. Short loops, no funnels. Want to see it from your side? Read [For Founders](https://gradpipe.com/for-founders) or [For Engineers](https://gradpipe.com/for-engineers). --- # For Founders You're not short on résumés. You're short on engineers who have actually built the hard thing you need built. GradPipe finds those people and introduces you directly. ## The offer | | | | --- | --- | | **From** | a resume pile | | **To** | three real engineers | We hire by craft, not by credential. - **We review GitHub, portfolio, résumé, and LinkedIn** for every candidate, so you meet people whose work we've actually read. - **Three matches per role, not fifty.** A small, considered set you can talk to this week — not a pipeline to manage. - **First reply in a working day.** You tell us the role; we come back fast. - **Direct introductions.** No applicant tracking, no platform middleware, no nurture campaigns. You talk to the engineer; we get out of the way. - **Most placements close in under three weeks.** ## How it works for you 1. Tell us what you're building, what's hard about it, and who you've already tried to hire. 2. We read the role against our register and reply within a working day with three to five engineers whose proven work matches. 3. We introduce you directly. You decide who to move forward with. There's **no recruiter signup**. You [submit the role](#find-talent) and book a 15-minute slot with the founders in one go. ## Who's on the register The engineers we send have shipped at companies like Stripe, Google, Jane Street, Databricks, NVIDIA, and OpenAI; studied at IITs, ETH Zürich, CMU, and Yale; and published at venues like NeurIPS, ICML, and ICLR. We surface the ones whose proven work fits your specific problem — compiler and kernel engineers, distributed-systems builders, backend and search-infrastructure engineers, ML research scientists. ## What founders say > "GradPipe and Independence OS came up alongside each other. When I was > building a healthcare AI company from scratch, they opened the IIT Bombay and > IIT Kharagpur pipelines I couldn't have reached myself. Several of my founding > engineers came through them." > — **Ravinder Syal**, founder, Independence OS > "Muhammad and his team were very proactive and helped us find some really > strong SWE interns for Unsiloed AI." > — **Aman Mishra**, co-founder & CEO, Unsiloed AI (YC F25) ## Ready? [Find talent](#find-talent) or email contact@gradpipe.com. Read [how the scoring works](https://gradpipe.com/how-it-works) first if you want the technical detail. --- # For Engineers If you've written the hard parts — the kernel, the compiler pass, the migration that couldn't go down — you shouldn't have to grind job boards to be found. GradPipe brings the right roles to you. ## The offer | | | | --- | --- | | **From** | job board noise | | **To** | a role that fits | Be found by what you've shipped. - **Roles come to you** — not the other way around. - **Your work is the evaluation.** We read your GitHub, shipped systems, portfolio, résumé, and LinkedIn — your code is the interview. - **You decide if the work is interesting.** No spam, no funnel, no recruiter treadmill. We only bring you roles where the problem is real. ## How it works for you 1. Join the network and connect the work you're proud of — public repositories, shipped systems, projects. 2. We analyze your public code at the AST level and read the rest of your record to understand what you actually do well. 3. When a founder's role matches your proven work, we introduce you directly. You talk to the founder; we stay out of the way. ## Who joins The strongest engineers we work with rarely apply to job posts. They're maintaining libraries other people depend on, finding bugs deep in GCC and LLVM, designing systems at a scale most teams never touch, or publishing at NeurIPS and ICLR. You don't need an elite college on your résumé — you need work that proves you can build. ## Privacy Your profile isn't a public listing. We abstract identifying details in anything public-facing, and founders see your full record only when we make an introduction that you're part of. The work speaks; you stay in control of who sees your name. ## Join [Join as a builder](/candidates). It's free for engineers — we're paid by the companies that hire, only when they hire. --- # Pricing GradPipe is paid on results. You pay a **success fee only when you hire** an engineer we introduced. Nothing else. ## What you pay - **Success fee: 15% of first-year CTC** (total cost to company, not just base salary), due only when a candidate we introduced accepts an offer and starts. - **No retainer.** We don't charge to start a search. - **No subscription.** There's no platform seat or monthly fee. - **No charge for introductions that don't convert.** If you don't hire, you don't pay. ## What's included - Sourcing and screening against the whole builder record, including AST-level analysis of public code. - Three to five matched candidates per role, with our first set within a working day. - Direct introductions and support through to offer. - A replacement guarantee: if a placed engineer leaves within the guarantee window, we re-run the search at no additional fee. ## How it compares - **Versus contingency recruiters (typically 20–30%):** lower fee, and you're evaluating engineers whose actual code we've read rather than résumés a recruiter keyword-matched. - **Versus retained search:** no upfront retainer and no monthly cost — you only pay on a successful hire. - **Versus job boards and sourcing tools:** we don't sell you access to a database and leave you to filter it. We return a short, considered set. ## Getting started There's no recruiter signup. [Submit your role](#find-talent) and book a 15-minute slot with the founders in one go. Questions: contact@gradpipe.com. --- # Frequently asked questions Everything founders and engineers ask us most often about how GradPipe works, what it costs, and how we evaluate people. ### How does GradPipe score engineers? GradPipe reads the whole builder record rather than ranking people by credentials. For every engineer we review four sources together: public GitHub code (analyzed down to the syntax tree), shipped systems they built and ran in production, their portfolio, and the context from their résumé and LinkedIn. No single source decides anything — a clean repo with no users, a famous employer with no visible output, or a strong résumé with no shipped work each tells only part of the story. We combine them, attribute the hard parts to the person who actually wrote them, and benchmark against engineers who have already proven themselves at the highest level. ### What is AST-level code analysis and why does it matter? An abstract syntax tree (AST) is the structured representation of code that a compiler sees, beneath the surface text. Analyzing code at the AST level lets us look past cosmetic details — formatting, names a linter would fix — and read the engineering decisions underneath: how complexity is managed, how the design is decomposed, whether failure paths are handled, whether the work shows real concurrency or performance engineering. It matters because those decisions are what separate strong engineers from weak ones, and they're far harder to fake than a star count or a commit streak. ### Why don't you use LeetCode-style assessments? A timed algorithm puzzle measures how well someone performs a puzzle under a clock. In a world where language models can generate a passable solution to most of those puzzles in seconds, that signal is weaker than ever and it correlates poorly with the real work of designing and maintaining systems. The code an engineer has already written, reviewed, shipped, and kept running is much harder to fake and much closer to the job. We read that instead of administering a test. ### What kind of engineers are on the platform? The engineers we work with tend to be people who've written the hard parts: compiler and kernel engineers, distributed-systems and platform builders, backend and search-infrastructure engineers, and ML research scientists. They've shipped at companies like Stripe, Google, Jane Street, Databricks, NVIDIA, and OpenAI; studied at IITs, ETH Zürich, CMU, and Yale; and published at venues like NeurIPS, ICML, and ICLR. Many of them never apply to job posts, which is exactly why a founder can't reach them through a job board. ### How does GradPipe differ from LinkedIn Recruiter, Hired, or AngelList Talent? Those products give you access to a large database and leave you to filter it on keywords and self-reported titles. GradPipe is the opposite: we read the actual work — including public code at the AST level — and return a short, considered set of three to five engineers per role, with our first reply inside a working day. You're not buying search access and doing the screening yourself; you're getting a small number of people whose proven work we've already matched to your problem. ### How does pricing work? GradPipe is paid on results. You pay a success fee of 15% of first-year CTC (total cost to company, not just base salary), due only when an engineer we introduced accepts an offer and starts. There's no retainer, no subscription, and no charge for introductions that don't convert. Placements come with a replacement guarantee: if a placed engineer leaves within the guarantee window, we re-run the search at no additional fee. For engineers, GradPipe is free — we're paid by the companies that hire. ### What's the success-fee model and how is it different from retained search? A success fee means we're paid only when you actually hire someone we introduced. Retained search charges an upfront retainer plus staged fees regardless of outcome; contingency recruiters typically charge 20–30% on a hire. GradPipe charges 15% on success, with nothing upfront and nothing if you don't hire. That aligns our incentives with yours: we only make money when we've actually solved your hiring problem. ### What's the typical placement timeline? Fast. After you describe a role, we reply within a working day with three to five matched candidates. Because introductions are direct — no applicant tracking system, no platform middleware, no scheduled nurture campaigns — the loop between first conversation and offer is short. Most placements close in under three weeks. ### Which startups have hired through GradPipe? GradPipe has placed engineers at startups including Independence OS (a healthcare AI company that hired several founding engineers through us), Unsiloed AI (a YC F25 company that hired strong software-engineering interns), and Stellon Labs (which hired a founding engineer for low-level ML infrastructure). You can read the details on our case-study pages. ### Can engineers join if they're not from elite colleges? Yes. We evaluate proven work, not pedigree. Plenty of the strongest builders we work with never attended a brand-name school — what they have is code they wrote and systems they shipped. Because our analysis reads the work directly, an engineer without an elite-college line on their résumé competes on equal footing with one who has it. The bar is the work. ### How does GradPipe protect engineer privacy? An engineer's profile is not a public listing. In anything public-facing we abstract identifying details — the work is shown, the name and uniquely identifying specifics are not. A founder sees an engineer's full record only when we make an introduction that the engineer is part of. Engineers stay in control of who learns their name and when. ### What is Craft Score? Craft Score is GradPipe's way of expressing how strong an engineer's work is, derived from the AST-level analysis and the rest of the builder record. Rather than a single opaque number from a timed test, it reflects the engineering decisions visible in real, shipped code — benchmarked against engineers who have already proven themselves at the highest level. We disclose the categories of signal that feed it; the exact weights and thresholds are proprietary. ### Does GradPipe work with companies outside the startup world? Our focus is startups hiring senior and founding engineers — the situations where getting the first few technical hires right matters most and where a short, high-signal set of candidates beats a large pipeline. If you're a startup building something genuinely hard and need engineers who've done that kind of work before, that's exactly who we serve. ### How do I get started? If you're hiring, there's no recruiter signup — submit your role and book a 15-minute slot with the founders in one go. If you're an engineer, join the network and connect the work you're proud of. Either way you can email contact@gradpipe.com. --- # How Independence OS hired founding engineers through GradPipe **Company:** Independence OS — a healthcare AI company **Role:** founding engineers **Outcome:** several founding engineers placed via pipelines the founder couldn't reach alone ## The situation Independence OS was being built from scratch in healthcare AI — a domain where the first engineering hires set the technical trajectory of the entire company. Founding-engineer hires like these are the hardest a founder makes: you need people who can build hard systems with very little scaffolding, and the strongest of them are almost never browsing job boards. ## What was hard The talent the founder needed sat inside networks he had no direct line into — specifically the IIT Bombay and IIT Kharagpur engineering pipelines, where a lot of the strongest early-career systems and ML engineers come from. Reaching that caliber of person, early, before they were locked up elsewhere, was the bottleneck. ## How GradPipe helped GradPipe opened those pipelines directly. Because we evaluate engineers on the work they can prove — public code read at the AST level, shipped systems, and the rest of the builder record — we were able to surface people whose demonstrated work fit a hard, early-stage healthcare AI problem, and introduce them to the founder directly rather than routing them through a funnel. ## In the founder's words > "GradPipe and Independence OS came up alongside each other. When I was > building a healthcare AI company from scratch, they opened the IIT Bombay and > IIT Kharagpur pipelines I couldn't have reached myself. Several of my founding > engineers came through them. Early access to that caliber of talent was > decisive." > > — **Ravinder Syal**, founder, Independence OS ## The takeaway For a founder building something hard, the constraint is rarely the number of candidates — it's reaching the few who have actually done the hard thing before, early enough to hire them. That's the gap GradPipe closes. Hiring founding engineers? [Find talent](#find-talent). --- # How Unsiloed AI (YC F25) found strong SWE interns through GradPipe **Company:** Unsiloed AI — YC F25 **Role:** software-engineering interns **Outcome:** strong SWE interns sourced through a proactive, founder-led process ## The situation Unsiloed AI, a YC F25 company, needed software-engineering interns who could contribute quickly. For an early-stage startup, an intern isn't a side project — the right intern ships real work, and the wrong one is a drain on a tiny team's time. The bar was high and the search needed to be fast. ## What was hard Intern hiring usually means wading through a large volume of applications with thin signal: coursework, a few class projects, and self-reported skills. Finding the few candidates who could actually build, without spending the team's scarce hours screening, was the challenge. ## How GradPipe helped GradPipe evaluated candidates on the work they could prove — public code and shipped projects read directly — and returned a short, high-signal set rather than a pile to filter. The process was hands-on: the founders stayed close to the search and moved fast on the strongest matches. ## In the founder's words > "Muhammad and his team were very proactive and helped us find some really > strong SWE interns for Unsiloed AI." > > — **Aman Mishra**, co-founder & CEO, Unsiloed AI (YC F25) ## The takeaway Even for internships, reading the actual work beats filtering résumés. A small, considered set of candidates who can demonstrably build saves an early team the thing it has least of: time. Hiring interns or engineers? [Find talent](#find-talent). --- # How Stellon Labs hired a founding engineer through GradPipe **Company:** Stellon Labs — low-level infrastructure for machine learning **Role:** founding engineer **Outcome:** a founding engineer sourced and placed in a matter of weeks ## The situation Stellon Labs works close to the metal — the kind of low-level infrastructure that machine-learning systems run on. A founding engineer for work like this has to be comfortable in a regime most engineers never touch: systems programming, performance constraints, and the unglamorous correctness work that infrastructure demands. ## What was hard The pool of engineers who can genuinely do kernel- and systems-level infrastructure work is small, and the ones who can are rarely on the market in any visible way. Conventional sourcing — keyword-matching résumés or scanning a recruiting database — surfaces people who *say* they do systems work far more often than people who demonstrably *have*. ## How GradPipe helped This is exactly the case our methodology is built for. By reading public code at the AST level, we can see who has actually written the hard, low-level parts — not just who lists the right keywords. We identified an engineer whose proven work matched the infrastructure problem and introduced them to the founders directly, closing the placement in a matter of weeks. ## The takeaway The harder and more specialized the role, the more a credential-based search fails and the more reading the actual work pays off. For deep infrastructure hires, demonstrated work is the only reliable signal. Hiring for hard systems work? [Find talent](#find-talent).