P

Pangram Labs Tech Stack

AI detection platform with peer-reviewed accuracy for written content

Technology, Information and Internet Brooklyn, NY 2–10 employees Privately Held

Pangram Labs builds an AI detection model claiming 99.98% accuracy, backed by peer-reviewed research. The tech stack reveals a heavy ML-infrastructure focus: PyTorch, CUDA, DeepSpeed, Ray, and vLLM for model training and inference, paired with Apache Spark, Beam, and Airflow for data pipelines. Active projects span synthetic dataset creation, LLM inference scaling, and a synthetic text generation pipeline—indicating the company is building both the detection capability and the training data to sustain it. Hiring velocity is accelerating across engineering and research roles, suggesting they're scaling inference capacity and model development in parallel.

Tech Stack 27 technologies

Core StackPyTorch Gusto Ashby QuickBooks Python Apache Spark Apache Airflow AWS React Next.js Django Ruby on Rails Mixpanel Amplitude NVIDIA CUDA DeepSpeed Ray vLLM Beam GCP PostHog iOS Android Canvas Blackboard Brightspace

What Pangram Labs Is Building

◆Challenges

Operational gaps
Ad hoc operational tasks
Scaling inference pipelines
User acquisition and engagement
Navigating legacy systems
Reverse engineering partner behavior
Poor documentation

▲Active Projects

Core ai detection model
Synthetic dataset creation at scale
Contract lifecycle management
Internal tools development
Process improvement for scaling
Synthetic text generation pipeline
Llm inference pipeline
Growth experiments
Landing pages
Referral loops

Hiring Activity

Accelerating6 roles · 6 in 30d

Department

Engineering

3

Ops

1

Research

1

Seniority

Mid

3

Lead

1

Senior

1

Company intelligence

Find more companies like Pangram Labs by tech stack, pain points and active projects

Get started free

About Pangram Labs

Pangram Labs operates a machine learning platform for detecting AI-generated writing, targeting educators, publishers, and content platforms. The company is based in Brooklyn and currently operates at 2–10 employees. Their immediate technical focus centers on three areas: improving the core detection model, scaling inference pipelines for user-facing throughput, and expanding synthetic training datasets. They're also building internal tooling and experimenting with growth loops. Documented pain points include operational scaling, inference pipeline bottlenecks, and user acquisition—typical constraints for early-stage ML platforms balancing model quality with product adoption.

HeadquartersBrooklyn, NY

Company Size2–10 employees

Hiring MarketsUnited States

Frequently Asked Questions

What is Pangram Labs' claimed accuracy for AI detection?

Pangram claims 99.98% accuracy. Their approach is grounded in peer-reviewed research and focuses on helping users understand the origin of written content rather than simple binary classification.

What machine learning infrastructure does Pangram Labs use?

The stack includes PyTorch, CUDA, and DeepSpeed for model training; Ray, vLLM, and Apache Airflow for inference and data orchestration; and AWS/GCP for cloud compute. This setup supports both model development and production inference scaling.

Similar Companies in Technology, Information and Internet

Other companies in the same industry, closest in size

V

VIHU

Technology, Information and Internet

C

Candidata ai

Technology, Information and Internet

K

Kanz

Technology, Information and Internet