Atul Pal

AI and Deep Learning professional with 4.5+ years of experience in Generative AI, NLP, and computer vision, specializing in scalable AI solutions and model optimization. Proficient in Generative Models, Diffusion Models, Transformers, Image Synthesis, and Edge AI, delivering innovative applications across medical imaging, security, and automation. Passionate about creating high-impact AI solutions that enhance efficiency and empower industries.

Senior Data Scientist at Kipi.ai (2026-03 – Present)

Responsible for designing and deploying end-to-end machine learning solutions for predictive analytics, including escalation prediction systems. Leveraged transformer-based models such as BERT, along with LLMs and vision-language models (e.g., CLIP), to enable multi-modal understanding across text, documents, and images. Built hybrid pipelines combining deep learning embeddings with LightGBM/XGBoost for structured and unstructured data.

Developed scalable data pipelines and real-time inference systems using FastAPI and Docker, and implemented explainability, monitoring, and automated retraining pipelines to ensure robust, production-grade performance.

• Escalation Prediction System – Developed an NLP-driven system to predict high-risk customer support ticket escalations using BERT-based text understanding and behavioral features such as sentiment trends, response latency, and interaction patterns. Combined transformer embeddings with gradient boosting models to improve prediction accuracy, and deployed the solution as a real-time service integrated with ticketing platforms, enabling proactive intervention and improved SLA adherence.

Data Scientist II at Endovision Tech Pvt. Ltd (2023-07 – 2026-03)

Responsible for designing, developing, implementing, fine-tuning, and optimizing deep learning and computer vision models for object detection, recognition, segmentation, and tracking, using large-scale medical imaging datasets as well as PDFs, text files, machine-printed, and handwritten documents.

Leveraged

Generative AI techniques, including GANs, Diffusion Models, and Large Language Models (LLMs), and Vision-Language Models (VLMs), for synthetic data generation, document understanding, and clinical report automation, improving model accuracy and generalization. Designed and deployed real-world AI applications on edge devices and cloud servers, ensuring low-latency, scalable, production-ready solutions. Bridged research and practical deployment to deliver high-impact AI solutions across medical imaging, document intelligence, and automation, effectively solving complex problems.

• Synthetic Data Generation – Designed and implemented GAN and diffusion models to synthesize normal and abnormal stomach images, enhancing dataset diversity and improving downstream anomaly detection and segmentation robustness.
• Clinical Report Generation – Built an automated clinical report generation pipeline using LLaMA and BERT, fine-tuned on medical text to improve factual accuracy and reduce manual reporting effort.
• Detection & Segmentation Model – Designed and deployed YOLO and U-Net–based models for detection and segmentation of gastric lesions (polyps, ulcers, EGC), enabling accurate anatomical landmark localization in endoscopic imaging, improving model robustness through synthetic data augmentation.
• Image Processing & Classification – Designed and deployed computer vision models and algorithms for medical imaging, combining classical image processing and machine learning-based classification for feature extraction, pattern recognition, and image manipulation. Processed and analyzed large-scale imaging datasets to enhance diagnosis accuracy and support clinical decision-making.
• AI Annotation Platform – Engineered a desktop-based annotation tool with semi-automated labeling using pre-trained models, accelerating data generation and enabling iterative training and inference workflows.
• AI Deployment Platform – Developed a reusable, configurable ML platform for rapid development and deployment of applications, supporting scalable execution across edge using Dockerized architecture.

AI/ML Engineer at V-labs.ai (2022-05 – 2023-07)

Designed and developed models for computer vision and NLP tasks, including object detection, named entity recognition (NER), text extraction, and database vector search, and performed model optimization for efficient, high-performance inference. Worked with large-scale datasets, including RGB and thermal images, text images, PDFs, and scanned documents, to extract meaningful features and insights. Deployed models on cloud infrastructure to handle high-throughput inference and scale processing for large volumes of data in real-time.

• Detection Model – Designed and deployed a YOLO-based object detection model for RGB and thermal video feeds, enabling accurate real-time detection in diverse environments.
• Document Intelligence – Implemented OCR-based text extraction for handwritten and machine-printed documents and built BERT-based NER models to accurately identify and classify key entities. Integrated database vector search for efficient retrieval and deployed models on cloud infrastructure to handle high-throughput inference and scale processing for large volumes of data in real-time.

Data Analyst at Cointab Software Pvt. Ltd (2021-10 – 2022-05)

Designed and developed solutions for CV and NLP applications. Built OCR for extracting structured data from unstructured documents, including PDFs and images. Developed ML models for signature detection, document classification, and anomaly detection, improving data accuracy, document security, and operational efficiency.

• Detection Model – Developed ML models for automated signature detection and verification.
• Document Intelligence – Implemented an OCR-based system to extract structured data from scanned invoices and convert it into a predefined structure using NER models for accurate entity recognition.

Freelance Machine Learning Engineer | AI, Computer Vision & Image Generation Specialist at Freelance (2022-05 – Present)

Freelance Machine Learning Engineer with 1.5 years of experience developing and deploying AI solutions in computer vision, deep learning, Large Language Models (LLMs), and image generation. Built custom applications including camera calibration tools, text extraction systems, automated attendance systems, data generation pipelines, and AI-powered image synthesis models. Skilled in Python, TensorFlow, PyTorch, OpenCV, and other frameworks, with expertise in Generative AI, GANs.

Experienced in model optimization, hyperparameter tuning, and deployment on edge devices and cloud servers for high-performance, scalable AI systems. Committed to delivering innovative, real-world AI solutions that automate processes, enhance decision-making, and generate synthetic data for improved analytics.

M.TECH – IIT KANPUR (2021)

B.TECH – AKTU LUCKNOW (2019)