Ishan M.

Key Skills

Software

Google Cloud Vertex AI

Data Annotation Tech

Labelbox

Scale AI

Appen

CVAT

Mindrift

Top Subject Matter

Technology – AI Training, Computer Vision & LLMs

Gaming & Software – UI/UX Screenshots & Dynamic Environments

Automotive – Object Detection & Segmentation

Top Data Types

Text

Document

Image

Top Task Types

Point/Key Point

Polyline

Text Generation

Question Answering

Text Summarization

Polygon

RLHF

Fine-tuning

Evaluation/Rating

Transcription

Computer Programming/Coding

Data Collection

Prompt + Response Writing (SFT)

Freelancer Overview

I am a Cloud and AI Developer and the founder of two development agencies, specializing in the intersection of Computer Vision, Data Labeling, Agentic AI, and Full-Stack Engineering. With a technical background in B.Tech (AI & ML), I bridge the gap between high-level architectural design and the granular data requirements necessary for state-of-the-art model training. My expertise includes: Data Annotation & RLHF: Proven track record in high-precision polygon segmentation, specifically for complex UI screenshots, gaming data, and automotive environments using tools like Labelbox. AI Development: Experience building autonomous agents and custom LLM-based systems, including multi-agent orchestration and gesture-controlled neural interfaces. Software Engineering: Proficient in Python, cloud architecture, and hardware-integrated robotics (Arduino/ESP32), allowing me to understand the full lifecycle of AI-integrated products. I am committed to delivering high-quality, spatially accurate datasets and robust software solutions under tight deadlines, ensuring that the foundational data drives superior model performance.

Labeling Experience

Multi-Domain Advanced Data Synthesis & Vision Model Training

ImagePolygon

Led high-precision annotation initiatives across diverse sectors, including autonomous vehicle perception and large-scale retail categorization. I specialize in complex polygon segmentation and multi-layer object detection, focusing on edge-case accuracy and structural integrity of datasets. My work ensures that the spatial data used for training deep learning models meets the highest industry standards for reliability and reasoning.

2025 - Present

RLHF & LLM Alignment for Reasoning and Factuality

TextRLHF

Contributed to the alignment of advanced Large Language Models (LLMs) through Reinforcement Learning from Human Feedback (RLHF). I specialized in evaluating and ranking model responses based on truthfulness, helpfulness, and safety. My work involved identifying subtle logical fallacies, correcting technical hallucinations, and refining the model's ability to follow complex, multi-step instructions in both English and Bengali.

2025 - 2026

Education