Sonu Suman

About Sonu Suman

With experience in data refinement and ML model training at Innodata, I have developed a…

Experience

1. Senior Associate - Ai Ml Nlp & Computer Vision

   Innodata Inc.

   Sep 2024 — Present · Noida, IN

   Specialization: High-precision data extraction, NLP, and computer vision for scalable AI solutions.Core Responsibilities:1. Led end-to-end AI model training/deployment for 3 enterprise projects, boosting accuracy by 20-30%.2. Automated extraction of structured data from PDFs, MMD files, images, and Excel sheets using Python, pandas, and NumPy.3. Ensured data integrity for AI/Generative AI training through strict quality control and cross-functional collaboration.4. Streamlined workflows using NLP techniques (text cleaning, tokenization, NER) to extract insights from unstructured data.Key Projects & Achievements:1. AI Model Training and Data Refinement for Q&A Extractiona. Optimized AI models using RWSW/LaTeX for MMD/PDF analysis, improving training data quality by 35%.b. Built NLP pipelines for Q&A extraction:i. Automated text cleaning, tokenization, NER, reducing manual effort by 40%.ii. Refined 50K+ data samples with data engineers, enhancing model accuracy by 25%.2. Google Sunroof Solar System - AI Model for Solar Feasibility Prediction (Client: Google)a. Trained ML models on aerial imagery, achieving 95% accuracy in solar panel viability prediction.b. Automated labeling/preprocessing with Python + DataCompute, cutting manual effort by 30%.c. Integrated IoT sensors + Google Cloud for real-time predictions, enabling scalable energy solutions.3. AI-Driven Precision Agriculture: Street-View Image Analysis for Crop Classification (Anthrokrishi Project)a. Classified crops via CNN models on street-view images (93% accuracy).b. Resolved edge cases (occlusion, ambiguous land types) through advanced preprocessing.c. Identified crop stages, aiding farmer decision-making.Technical Impact:1. Scaled 3 AI solutions (including Q&A systems refined via reinforcement learning), impacting 1M+ users in agriculture and renewable energy sectors.2. Enhanced AI responsiveness by 15-20% through iterative human feedback integration in Q&A pipelines.