Founding Member of Technical Staff
Vectoral
Founding Member of Technical Staff
Vectoral
Detects and blocks LLM token abuse
Vectoral is a B2B security company protecting LLM API traffic from fraud and abuse, including token resellers, account farms, denial-of-wallet attacks, automated scraping, CSAM, and model distillation.
We’re already working with some of the largest names in inference and are looking for exceptional engineers to help us build and scale the security layer for AI infrastructure.
Own whether we catch the fraud. Everything that goes into making a verdict. The signals, rules, trained models, thresholds, and evidence that it works.
The detection algorithm itself: a rule-and-group engine in Go that produces every live verdict. Signal groups like machine-paced behavior, birth cohort, probing, resource shape, value extraction, datacenter origin, and account risk. Each with activation logic, per-customer presets, calibration, and cold-start handling for accounts we've barely seen.
A LightGBM training pipeline (Python, psycopg, scikit-learn) that pulls labels and feature snapshots per customer, fits a binary classifier, Platt-calibrates on a temporally held-out split, and writes a candidate model.
A model-replay worker that scores stored snapshots offline, allowing new changes to be evaluated against real historical traffic before affecting live verdicts.
An adversarial red-team harness where a blind LLM adversary plays an abuser attacking a customer's product.
We want to build an exceptional team, and we're willing to go above and beyond to get the right people.
Competitive salary and equity
Health, dental, and vision insurance
Unlimited paid time off
Flexible work hours
Someone who is an applied ML or detection/anti-abuse practitioner. They ship the model and the code that runs it in production, on a latency-sensitive path
Built fraud, abuse, anti-bot, risk, or security detection systems that operated against live adversaries, and can discuss how these systems were evaded and the actions taken in response
Fluent in gradient boosting, calibration, temporal train/test splitting, and precision/recall tradeoffs under severe class imbalance; understands why AUC alone is insufficient to argue a model is ready
Strong Python; can read and write Go, or will get there fast.
Build from day one on a product making a real impact with active customers.
Founding engineering role: shape product, architecture, and company culture from the ground up.
Significant technical ownership and direct influence over what we build and how we build it.
Work with a tight, highly skilled team on problems where robust systems matter and where your work directly contributes to customer security and satisfaction.
If you’re excited to build products that stand up to adversaries, enjoy shipping code that matters, and want to join an early team where your impact is outsized, we want to talk.
Vectoral protects large language model endpoints from token reselling and API abuse. It monitors incoming requests to detect and block reseller traffic before tokens are spent. The platform uses behavioral fingerprints, request-graph clustering, and abuse-network intelligence to score and categorize each request as blocked, challenged, or clean, and logs the trigger signal. It lets teams rank users by spend, block abusive accounts, and set actions such as throttling, key rotation, or refusing calls inline. This helps providers prevent unauthorized use and minimize API costs. Vectoral aims to reduce fraud and abuse in generative AI services by stopping credential leaks and token resale at the gateway.
Explore more opportunities at Vectoral