Harrison Jilani - AI Data Trainer & Analyst · LLM Evaluation · RLHF · Computer Vision · 3D LiDAR

Key Skills

Software

CVAT

Labelbox

SuperAnnotate

LabelImg

Hasty

Scale AI

Sama

Appen

Remotasks

Roboflow

V7 Labs

Toloka

Mindrift

CloudFactory

Micro1

Mercor

OneForma

Surge AI

Data Annotation Tech

Top Subject Matter

LLM Evaluation — English (RLHF, SFT, Safety)

Computer Vision and Image Annotation

3D LiDAR Data Annotation (Autonomous Vehicles)

Top Data Types

Image

Video

3D Sensor

Top Task Types

Bounding Box

Classification

Evaluation/Rating

Prompt + Response Writing (SFT)

RLHF

Fine-tuning

Object Detection

Red Teaming

Text Generation

Segmentation

Polygon

Transcription

Computer Programming/Coding

Freelancer Overview

AI Data and Research Specialist with 6+ years of hands-on experience across the full AI and machine learning pipeline, from raw data collection and multimodal annotation to LLM evaluation, RLHF, and QA framework design. I have evaluated Large Language Model (LLM) outputs for factual accuracy, instruction-following, reasoning quality, and human preference alignment, directly supporting fine-tuning and safety pipelines. My work spans hallucination detection, red teaming, chain-of-thought assessment, and Supervised Fine-Tuning (SFT), with every finding backed by detailed written justifications used directly by model developers. On the data side, I have labeled and validated multimodal datasets across text, image, audio, video, and 3D LiDAR point clouds proficiently using Labelbox, CVAT, Superannotate, and LabelMe. I specialise in bounding boxes, semantic and instance segmentation, ASR transcription, NLP annotation, and autonomous vehicle datasets. As a team leader, I have onboarded and coached 50+ contributors, redesigned QA workflows that reduced error detection time by 30%, improved dataset precision by 15%, and cut client rejection rates by 12%. I bring the same rigour to individual contributor roles, consistently meeting SLAs and delivering above quality thresholds. I am equally comfortable operating independently on tight deadlines or embedded within distributed global teams, and I am committed to helping clients build cleaner, safer, and more capable AI systems.

ExpertSwahiliEnglish

Labeling Experience

Large-Scale Multimodal Image Annotation & Data Quality for SFT Pipelines

ImageBounding Box

At Fuzu Kenya, I review and approve 10,000+ images per week to enforce labeling consistency and dataset integrity for Supervised Fine-Tuning (SFT) pipelines and ML model development workflows. My responsibilities include applying structured taxonomies and classification schemas to ensure image datasets are correctly tagged and pipeline-ready, detecting and correcting mislabeled samples, ambiguous classifications, and edge cases to reduce noise in training data. I also flag policy-violating content to uphold safety and content moderation standards, and deliver written feedback to technical teams on recurring tool and workflow issues. Tools used include Labelbox, Superannotate, CVAT, and LabelMe. This role demands high accuracy at speed — maintaining consistent quality across large daily volumes with zero tolerance for dataset contamination.

2026 - Present

3D LiDAR Point Cloud Annotation for Autonomous Vehicle AI Systems

3D SensorObject Detection

At CloudFactory, I specialised in annotating 2D and 3D LiDAR point cloud datasets used to train and validate AI algorithms for autonomous vehicle and spatial AI systems. My work involved identifying and classifying objects — including vehicles, pedestrians, cyclists, and road markings within LiDAR point clouds using bounding boxes, cuboids, instance segmentation, and semantic classification. I generated ground truth datasets to strict quality and consistency standards, closely following project guidelines and promptly resolving data anomalies and labeling conflicts to strengthen inter-annotator agreement. I led daily operations for a team of 10+ contributors across LiDAR, NLP, and ASR pipelines, managing task distribution, output reviews, and deadline adherence. I was awarded Team Lead of the Quarter for sustaining the highest output standards on one of the platform's highest-volume data pipelines.

2024 - 2025

LLM Output Evaluation, RLHF & Safety Assessment for Generative AI Models

TextRLHF

Over 5 years at Sama Kenya, I evaluated thousands of LLM-generated responses using structured rubrics covering factual accuracy, instruction-following, reasoning quality, and human preference alignment directly supporting RLHF and Supervised Fine-Tuning (SFT) pipelines for generative AI models. My core responsibilities included detecting hallucinations, unsafe outputs, and prompt misalignments, and documenting every finding with detailed written justifications passed directly to model development teams. I also conducted red teaming exercises to surface failure modes and edge cases in model behaviour. As the project scaled, I designed QA frameworks and performance dashboards that reduced error detection time by 30%, improved dataset precision by 15%, and lowered client rejection rates by 12%. I onboarded and mentored 50+ evaluators in rubric methodology, safety standards, and calibration, and advanced from an entry-level contributor to a Senior QA Analyst and Subject Matter Expert throughout the engagement.

2018 - 2023

Education

S

St. Paul's University

Bachelor of Science in Computer Science, Computer Science

Bachelor of Science in Computer Science

2018 - 2023

Work History

F

Fuzu Kenya

AI Data Quality Specialist | Multimodal Annotation Analyst

Nairobi

2025 - Present

I

Impact Outsourcing Ltd

Research Analyst | Compliance & Vessel Intelligence

Nairobi

2024 - 2024