Open to AI data & tooling work

Adnan Ahmed

Founding Software Engineer at human depth.
Human Data Expert at micro1.

I build the audit, QA and fraud-detection tooling behind large-scale AI training data pipelines — and manage the people who run them.

Based: Nakuru, Kenya
Working: Remote
Focus: Human data · QA · Internal tooling

Book a call adnan@humandepth.org

0%+

Sustained quality score

AI Expert Generalist, micro1

Internal tools shipped

Audit, QA, fraud detection & reporting

0 mo

Annotator to pipeline lead

Jan 2026 → Feb 2026

Concurrent frontier-AI roles

human depth · micro1

02 — Experience

Two roles, both current, both in the human layer of AI.

micro1 supplies expert human data to frontier AI labs. human depth is building the platform for the people who do that work. I'm inside both.

micro1

Current

Human data engine for frontier AI labs

Remote — Palo Alto, CA

Annotator to running the pipeline, in a month.

Human Data Expert

Feb 2026 — Present

Manage a team of generalists and Human Data Managers supporting a large-scale robotics training data pipeline, owning quality review, fraud detection, onboarding and payments operations end to end.
Designed and built FrameCapture, an internal desktop audit tool with a multi-threaded browser-automation architecture, to verify submission integrity across the annotation pipeline.
Built UIAR, a UI augmentation layer adding task-tracking identifiers to internal project boards, improving pipeline visibility for reviewers and leads.
Designed a QA reporting framework combining session replay, activity logging and network capture into a single exportable audit report supporting quality and fraud-review decisions.
Built a biometric fraud-detection system using computer-vision tracking and anomaly-detection modelling to flag suspicious submission behaviour at scale.
Built an automated pipeline-activity reporting system with quality scoring and refreshing visualizations to track team and pipeline performance.
Partnered directly with a Strategic Project Lead on an AI-Lab-facing engagement, contributing to the design, build and optimization of the labeling tools supporting the project.

Promoted after one month

AI Expert Generalist

Jan 2026 — Feb 2026

Delivered AI training data annotations with a sustained quality score above 95%, maintaining strict adherence to evolving project guidelines and workflows.
Coordinated with peer annotators to share workflow improvements, contributing to team-wide submission quality and approval rates.

human depth

Current

Pre-launch platform for the human layer of AI

Remote — London, UK

Founding engineer on a platform for the people behind the models.

Founding Software Engineer

Jul 2026 — Present

Building a professional community platform for AI training, annotation and operations professionals, working directly alongside the COO to define the platform's technical foundation.
Co-designing a role-progression system to help human-in-the-loop professionals map career paths across the industry, and a multi-grader certification framework for verifying skill quality.
Contributing to competitive positioning strategy, differentiating the platform from traditional HR and referral models around assisted candidate positioning and tailoring.

03 — Tooling

Five internal tools, built because the pipeline needed them.

Audit, QA, fraud detection and reporting systems built inside micro1's robotics training data pipeline. Not side projects — infrastructure the work depended on.

FrameCapture

Internal desktop audit tool

Problem: Submission integrity across a large-scale annotation pipeline had to be verified at a volume manual spot-checking could not cover.
Build: A desktop audit tool built on a multi-threaded browser-automation architecture, driving verification passes concurrently rather than serially.
Effect: Verifies submission integrity across the annotation pipeline.

UIAR

UI augmentation layer

Problem: Internal project boards carried no task-tracking identifiers, leaving reviewers and leads without a reliable way to follow work through the pipeline.
Build: An augmentation layer that injects task-tracking identifiers directly into the existing boards, rather than asking the team to adopt a parallel system.
Effect: Improved pipeline visibility for reviewers and leads.

QA Reporting Framework

Audit reporting system

Problem: Quality and fraud-review decisions depended on signals scattered across separate systems, with no single artifact a reviewer could act on.
Build: A framework combining session replay, activity logging and network capture into one exportable audit report.
Effect: Supports quality and fraud-review decisions from a single, self-contained record.

Biometric fraud detection

Computer vision + anomaly detection

Problem: Suspicious submission behaviour needed catching at a scale and consistency no human review rota can sustain.
Build: A detection system pairing computer-vision tracking with anomaly-detection modelling over submission behaviour.
Effect: Flags suspicious submission behaviour at scale.

Pipeline activity reporting

Automated reporting & scoring

Problem: Team and pipeline performance had to be tracked continuously, without a standing manual collation cost every reporting cycle.
Build: An automated activity-reporting system with quality scoring and self-refreshing visualizations.
Effect: Tracks team and pipeline performance on a live basis.

Also partnered directly with a Strategic Project Lead on an AI-Lab-facing engagement, contributing to the design, build and optimization of the labeling tools supporting the project.

Python
Google Apps Script
Chrome Extensions
QML & Qt
MediaPipe
LiDAR Data
Quality Assurance
Fraud Detection
Pipeline Operations
KPI Reporting

04 — Capabilities

Three sides of the same job.

Running a data pipeline means owning the operations, the people and the tools. These are the parts I actually do, not a wishlist.

01Operations
02Leadership
03Engineering

Operations

Running the pipeline, not just working in it.

Data Pipeline Operations & Management
Quality Assurance & Fraud Detection
Onboarding & Payments Operations
KPI Definition & Reporting
Process Optimization

Leadership

Setting the quality bar and holding it.

Team Leadership & People Management
Cross-Functional Project Management
Client & Stakeholder Communication

Engineering

Building the tools the work needed.

Python
Google Apps Script
Chrome Extension Development
QML & Qt
Computer Vision (MediaPipe)
LiDAR Data

05 — Credentials

Verifiable, not decorative.

Every completed badge links straight to its Credly record. Click through and check.

Completed

In progress

FCF & FCAFortinetOngoing
IT Support Specialist CourseworkProfessional certificateOngoing

Menengai High School — High School Diploma, Nakuru, 2020–2023

06 — Earlier

Before the pipeline.

Media, brand and IT work in Nakuru. Different domain, same habit of taking ownership of the whole thing.

Nov 2025 — Jan 2026
RedLabs Entertainment
Director
Directed freelance video editing, graphic design and digital media services, managing client communications to hold delivery quality consistent.
Jun 2025 — Oct 2025
DuoMotion
Co-Founder
Co-founded a media and brand studio; designed the core brand identity, logo, social assets and client-facing documents, and produced video content for social marketing campaigns.
Apr 2025 — Sep 2025
Maya's Experience
Information Technology Consultant
Managed company technology infrastructure and digital presence, including website management and equipment recommendations; produced multimedia content for livestreaming and social channels; advised on local search visibility.
Oct 2024 — Nov 2024
Olivetrust Developers
Advertisement Coordinator & Assistant
Coordinated video marketing production for real estate advertising campaigns, including video editing and graphic design.

07 — Archive

Design work, kept for the record.

Self-directed brand exercises for fictional companies, plus studio and client work from 2024–25. It is not what I do now, and it is labelled as what it is.

5 projects · 27 piecesView archive

Silica Designs & Systems

2024

Self-directed brand exercise

A fictional semiconductor and device company with a Linux-derived ecosystem. Identity, sub-brands and product mockups, briefed and art-directed by me.

AUDE

2024

Self-directed brand exercise

A fictional audio hardware brand. Wordmark, packaging direction and product imagery.

ChowTown

2024

Self-directed brand exercise

A fictional fast-food brand aimed at families. Logo, menu system and campaign posters.

DuoMotion

2025

Co-founded studio — brand system

Identity and collateral for the media studio I co-founded: logo, brand guide, business cards, social assets and client-facing documents. No longer operating.