H

Himanshu Parashar

About

Detail

Ajmer, Rajasthan, India
Contact Himanshu regarding: 
work
Full-time jobs

Timeline

work
Job
school
Education
folder
Project

Résumé

Jobs verified_user 0% verified
  • N
    Research Intern – Oral Cancer Histopathological Image Analysis
    National Institute of Technology (NIT) Jamshedpur
    Dec 2025 - Current (11 months)
    • Working on automated detection and segmentation of oral cancer from whole-slide histopathology images, with the goal of supporting early and accurate cancer diagnosis.
    • Exploring pre-trained CNN and vision transformer backbones for tissue-level feature extraction, and implementing segmentation models to distinguish tumour, stroma, and epithelial regions in oral tissue samples.
    • Handling real-world dataset challenges such as stain colour variation across hospitals, high-resolution image tiling, and class imbalance through appropriate preprocessing and augmentation strategies.
  • I
    Research Intern – EEG Signal Analysis for Autism Detection
    Indian Institute of Technology (IIT) Kharagpur
    Aug 2025 - Nov 2025 (4 months)
    • Investigated P300 brain signals as neurological markers for autism spectrum disorder (ASD), working with EEG recordings collected from 15 subjects across multiple sessions in a Brain–Computer Interface (BCI) setting.
    • Designed a lightweight hybrid CNN–Transformer model that combines spatial and temporal feature learning to detect P300 responses from 8-channel EEG data, achieving 96.64% test accuracy on a heavily imbalanced dataset.
    • Model generalised well across individual subjects and sessions, with consistent performance validated through subject-wise, session-wise, and cross-validation evaluations – some sessions achieving perfect classification scores.
  • I
    Deep Learning Research Intern – Human Activity Recognition
    Indian Institute of Information Technology, Design and Manufacturing (IIITDM), Kurnool
    May 2025 - Jul 2025 (3 months)
    • Proposed a hybrid WaveNet–BiLSTM–GRU architecture for recognising human activities from wearable sensor data, combining multi-scale temporal convolutions with sequential modelling to capture both short and long-range motion patterns.
    • Built a complete data processing pipeline for accelerometer and gyroscope time-series and evaluated the model on three standard benchmarks – UCI HAR, WISDM, and MHEALTH – achieving top accuracy of 99.86% on WISDM and strong results across all datasets.
    • Conducted thorough evaluation including confusion matrices, ROC curves, and per-class reports; manuscript prepared for journal submission.
  • D
    LLM Evaluator and benchmarking expert
    DuskerAi
    Jan 2025 - Apr 2026 (1 year 4 months)
    I Worked as a LLM Evaluator and benchmarking expert where I have to create task in different categories to stump the model
  • A
    AI Data Annotator
    Appen (Remote)
    Jan 2024 - Current (2 years 10 months)
    Performed data annotation, quality assurance, and multilingual evaluation for AI training pipelines.
  • O
    Language Model Trainer
    Outlier AI (Remote)
    Jan 2024 - Jan 2025 (1 year 1 month)
    Contributed to LLM fine-tuning through prompt engineering, multilingual evaluation, and preference ranking. Maintained 90%+ task accuracy.
Education verified_user 0% verified
  • C
    Integrated M.Sc in Computer Science
    Central University of Rajasthan (CURAJ), Ajmer, India
    Jan 2021 - Jan 2026 (5 years 1 month)
    CGPA: 7.44/10
  • J
    Higher Secondary (CBSE)
    Jawahar Navodaya Vidyalaya, Muzaffarpur, Bihar, India
    Jan 2020
    83%
Projects (professional or personal) verified_user 0% verified
  • Handshake AI
    I worked as a LLM Evaluator and Handshake AI
    Handshake AI
    Mar 2025 - Jun 2026 (1 year 4 months)
    I have to create task in different categories to stump the model and I worked on git hub fork rhe pr and push the work and wait to see the results and merged
  • Central University of Rajasthan
    DCFNet: Dual-Encoder Cross-Attention Fusion Network for 3D Liver Segmentation from T2-Weighted MRI
    Central University of Rajasthan
    Jan 2025 - Jan 2026 (1 year 1 month)
    • Designed DCFNet, a novel 3D segmentation network that uses two parallel encoders working together – one focusing on standard spatial features and the other capturing long-range context via 3D Coordinate Attention – fused through a Cross-Attention Fusion Module for adaptive liver boundary detection.
    • Achieved state-of-the-art liver segmentation on the CirrMRI600+ dataset (318 MRI scans), reaching 92.22% Dice score – a significant improvement over prior best (86.51%) and all baseline models including Attention U-Net, 3D U-Net, and V-Net.
    • Model predictions closely matched expert-annotated liver volumes and remained consistent across patients with varying liver shapes and disease stages, using a lightweight architecture of approx
  • C
    Continuous Fundus Image Quality Assessment (FIQA)
    Jan 2025
    • Developed a deep learning framework using an Inception-V3 backbone to automatically assess the quality of retinal fundus images, predicting continuous quality scores that reflect how usable an image is for clinical diagnosis.
    • Achieved strong correlation with human expert ratings, outperforming four state-of-the-art image quality baselines, making the system suitable for large-scale diabetic retinopathy screening programs.
  • L
    Lung Cancer Detection Using Deep Learning on DICOM CT Scans
    Jan 2024
    • Built and compared multiple deep learning models (CNN, ANN, LSTM, GRU) for classifying lung cancer types from over 40,000 DICOM CT scan images across 5 categories.
    • CNN achieved the best performance at 98.64% test accuracy, with full evaluation using confusion matrices, ROC-AUC curves, and per-class metrics.