Pre-seed → Series A · Early access open

Verified builder network. Real work. That you control.

Upswell runs on live Git telemetry, ranked peer attestations, and a double-blind trust layer you control. Built for the working engineer, not the platforms farming them.

Register for founder early accessCreate your build pass →
no scraped profiles·no keyword matches·no cold outreach
BUILD PASSBL–ML7734

ML Infra Engineer

8.4
Early-stageEx-high-growth
ROLE PASSCO–APZNPN

ML Infra Engineer

$200–275K
Remote-firstSeed-fundedExited
VELOCITY
INFRA
AI / LLM
Available now · synced 2m ago
92% FITrole match found
what we removed

Everything broken about the format. Removed.

0applications
0middlemen
0placement fees
0cold outreach
0scraped profiles
0unverified profiles
the noise problem

AI didn't fix hiring. It scaled the noise.

AI hiring tools scrape the same public data and repackage the keywords as a fit. Volume went up. Signal didn't. Upswell runs on live Git telemetry and peer attestations behind a double-blind trust layer: signal AI can't fabricate, and you never have to prove.

94% match
93% match
94% match
94% match
93% match
94% match
95% match
93% match
94% match
93% match
94% match
93% match
94% match
94% match
93% match
94% match
95% match
93% match
94% match
93% match
ranked on keyword overlap
If you're hiring

You're not drowning in candidates. You're drowning in candidates AI qualified for you.

It scored the words. You still have to find out if the engineer behind them can ship.

If you're building

Your public work is training data for tools you never opted into.

Scraped, ranked on stars and followers, and repackaged into a “profile” that isn't yours. Nothing about you exists on Upswell until you put it there.

no scraping·no synthetic profiles·no mass outreach
the pass

Two passes, built on one schema.

Generate a mock role pass →
Build Pass

Control your technical passport.

No form-filling, keyword optimization, or ghosting. Your Build Pass verifies your real-world technical signature straight from your Git log while keeping your identity fully private.

Automated Git Telemetry · Un-alterable

BUILD PASSBL–000000/active 6h ago
OPEN TO INTROS

ML Infra / Platform Engineer

Staff·7 yrs·$220k – $250k
8.9/10
SIGNAL SCORE
BUILDER STACK

Infra and AI heavy, light on frontend. Product sense is the gap, this role wants more of it than I have.

VELOCITYFRONTENDBACKENDINFRAAI/LLMPRODUCT
this builderrole envelope
GITHUB ACTIVITY52w
SKILLS● core  ○ familiar
PythonCUDAHuggingFaceGoRustTypeScriptKubernetesFastAPITerraformAirflow
CLAIMS6 claims · 14 attestations
01

Fine-tuning pipeline for a 7B-parameter LLM, shipped to production in six weeks, solo.

4 attestations
02

End-to-end model stack (training, eval, serving) owned solo across two teams.

3 attestations
03

Cut p99 inference latency from 800ms to 90ms via batching and kernel fusion.

3 attestations
04

llm-eval-harness, 800+ GitHub stars, still the internal source of truth.

2 attestations
05

Rebuilt the training data pipeline to cut a 14-hour job down to 40 minutes.

2 attestations
06

Ran the GPU cost review that took monthly spend from $48k to $19k.

AWAITING ATTESTATION2 vouches requested
BACKGROUND

Seven years, two AI startups, both pre-Series A. I owned ML infra at both: data pipelines, training, eval, and serving. No separate platform team, so I built and ran all of it. Cut inference latency 9x on a team of three. Wrote the eval harness both companies still use to gate model releases. Before that, three years on a platform team doing on-call for other people's services.

WHAT I WANT

Looking for a small team where I can take on the full path from experiment to production. Most of the hard work happens after training: serving, evals, cost, and keeping it up, and that's the part I'm good at. Happy to be the person who handles it so the rest of the team can stay on the model. Staff level, remote, seed or Series A.

IDENTITY
Founding engineerInfra / DevOpsML / AIEarly-stage (1–20)Open source
LOOKING FOR
StaffRemoteSeedSeries ADeveloper tools

Double-blind until you both opt in.

identity revealed on mutual accept · no recruiter in the loop

Request intro →
ROLE PASSCO–APZNPN/posted 4d ago·active 3h ago
INITIATING INTROS

ML Infrastructure Engineer

Staff·Founding hire·$200k – $260k
HIRING IMMEDIATELY
Remote
LOCATION
Series A
STAGE
3rd
INFRA HIRE
REQUIREMENTS4 hard gates · 2 preferred · ≥8.5 signal
01

Has shipped production ML training pipelines end to end, not just notebooks.

HARD GATE
02

Has operated low-latency model serving under real production load.

HARD GATE
03

Has built and owned an LLM evaluation harness other engineers rely on.

HARD GATE
04

Has debugged distributed training and GPU performance issues in production.

HARD GATE
05

Has owned infrastructure cost and capacity decisions, not just consumed them.

preferred
06

Has worked without a platform team to hand the tedious parts to.

preferred
WHAT YOU'D OWN

We build the model training infra layer for next-gen ML pipelines, from data ingestion through low-latency serving. You'd own training infra, eval harnesses, and the serving pipeline from experiment to production. Third ML infra hire, reporting to the CTO. Scope is wide and a fair amount of the tooling is still half-built.

FIRST 90 DAYS

Own the eval harness end to end. Ship the first fine-tuned model to production. Define the serving SLA with the CTO, and help hire the next two infra engineers.

COMPANY
Series AML / AI infraEarly-stage (1–20)Remote-firstTechnical founder
CONTEXT

Series A raised 8 months ago. Platform team scaling from 2 to 6 as usage climbs past capacity planning every quarter.

ROLE STACK

Infra and AI weighted, frontend out of scope. Needs some product judgment, you pick what gets built.

VELOCITYFRONTENDBACKENDINFRAAI/LLMPRODUCT
role envelopetop match
TEAM'S STACKweighted
PyTorch95
Python95
Kubernetes75
CUDA75
Airflow60
SKILLS● core  ○ familiar
PyTorchPythonKubernetesCUDAFastAPIAWSHuggingFaceTerraformAirflow

Double-blind until you both opt in.

company revealed on mutual accept · no placement fee

Request intro →
the score

Every number comes from your commits.

Connect GitHub once. Every number below is measured, not predicted.

Composite score
8.4/ 10

Recomputed on every push. Same inputs, same number, every time.

Last sync 2m ago
01Top 1%

Commit velocity

MedianYou
026.3

PR quality

MedianYou
035 / 6 axes

Stack depth

MedianYou
043 peers

Verified vouches

MedianYou
the workspaces

Two workspaces. One  zero-trust network.

upswellFounder
Search builders…
23AV
Overview
Pipeline
Builders
Passes
Settings
Role Pass live
Visible to 24 matched builders.
Strength100%

CO–APZNPN

Hiring now

ML Infrastructure Engineer · Senior · Remote · $200–275K

Edit passRequest intros
Active matches
24+5 wk
Avg signal
7.9/ 10
Intros requested
63 replied
Pass views · 7d
142+18%

Active matches

Ranked by signal · double-blind

Sort: Signal ▾
BuilderStackSignalStatus
BL–ML77ML / AI · Infra · Staff
8.4
Available
BL–KQ20Backend · Infra · Senior
7.9
~1 mo
BL–9FA3ML / AI · Data · Senior
7.4
Available
BL–7T18Infra · DevOps · Staff
7.1
~2 mo
BL–3C55Full-stack · ML · Senior
6.9
Available

Role Pass

What builders see

PREVIEW
CO–APZNPN
ML Infrastructure Engineer · Senior
Accelerator-backed, $3.2M raised, 8 on the team. Training-infra layer for next-gen ML pipelines.
VelocityFrontendBackendInfraAI/LLMProduct
Python · PyTorchExpert
Kubernetes · AWSProficient
upswellBuilder
Search roles…
RK
Available now
Overview
My Pass
Matches
Claims
Vouches
My Score
Settings

Welcome back, Riya

BL–ML7734

ML Infrastructure Engineer · Staff · Open to Seed–Series A

Edit passSet availability
Signal score
8.4+0.3
Incoming matches
72 new
Profile views · 7d
38founders
Commit streak
22d478 total

Incoming role matches

Founders who match your pass

Fit ▾
RoleCompanyCompFit
CO–APZNPNML Infra · Seed · Remote$200–275K92%
CO–7XR2Infra Platform · A · Hybrid$230–290K88%
CO–QL90Founding ML · Seed · Remote$190–240K81%
CO–BD14ML Systems · A · Onsite$240–300K76%
CO–5HM8Inference · Seed · Remote$200–250K73%

Build Pass

As founders see it

8.4SIGNAL
BL–ML7734
ML Infrastructure Engineer · Staff
Top 2% velocity478 commits
VelocityFrontendBackendInfraAI/LLMProduct
Shipped · p99 800ms→90ms✓ 4
OSS · llm-eval-harness✓ 3
how it works

Filter by telemetry. Connect by intent.

For foundersFor builders

Describe the stack, not the job

Stack, repo scale, and codebase complexity in under two minutes.

01Intake

Connect Git, read-only

Pick which repos are visible. Your source never leaves your machine.

Skip the whole screening layer

No resume triage, no agency shortlists, no AI-assembled lists.

02Analysis

Your pass builds itself

Radar, code footprint, and execution metrics generate from live data.

Vetting you can audit

Commit velocity, architectural depth, and peer-verified history.

03Validation

Get credit for work under NDA

The Vouch Loop turns unshareable work into verified peer attestations.

Read the radar, not a bio

Drawn straight from Git telemetry. Nothing to take on faith.

04Discovery

Filter by team mission

Real engineering expectations, team culture, and 90-day roadmaps.

05 · MUTUAL MATCH

Mutual opt-in drops the double-blind shield.

Contact details unlock only when both sides opt in.

match % = overlap between the tags a builder selects and the tags on a role. plain arithmetic, no model, nothing scraped.

the alternative

Every other tool was built for a different problem.

How they hire
LinkedIn
Agencies
Job boards
AI sourcing
Upswell
Cost to hire
$10–15K + per-click
20–30% salary
Per-post + seat
Seat + credits
$599 / slot · mo
Verified proof of work
–
–
–
–
✓
No cold outreach or spam
–
–
–
–
✓
No applications to write or screen
–
✓
–
–
✓
Peer attestations
Partial
–
–
–
✓
Time to first high-signal intro
Weeks
Weeks
Weeks
3–6 wks
Days
Early access open now

The new data plane for builders.

Connect in silence. Unlock by mutual intent.

Register for founder early accessCreate your build pass →