Devansh Gupta
Resume ↓

Mohali, Punjab, India

Devansh Gupta

AI Engineer Computer Vision / VLM / LLM / Medical Imaging

AI Engineer with 2+ years of experience in Vision-Language Models, Computer Vision, LLMs and medical imaging problems (1 year full-time and 1+ year in research). Built end-to-end solutions including training and deploying OCR models and a reward system for automated grading of answer sheets, alongside research in handwritten OCR, keypoint detection in medical imaging and deep learning models for signal processing. Skilled in VLM and LLM fine-tuning, scalable GPU deployments (vLLM) and full-stack AI products.

Email me Resume
Portrait of Devansh Gupta
About

Focus areas

AI/ML

Computer Vision OCR Layout Parsing LLMs RAG VLM/LLM Fine-tuning (SFT, RFT, GRPO) Explainable AI

Current Research

Handwritten OCR Keypoint Detection — document imaging Keypoint Detection — medical imaging
Experience

Where I've worked

  1. AI Engineer · Infutrix Technology

    Jun 2025 — Present · Mohali, India
    • Leading research on handwritten OCR and a reward-based grading system; trained a Qwen 3.5 9B model to perform handwritten OCR, map each answer to its question and assign rewards for scoring.
    • Researched character-level keypoint detection and space detection on line-level handwritten data.
    • Built an end-to-end answer sheet grading platform using CV–LLM pipelines with rubric-aware partial-credit grading, explainable AI feedback, and monitoring and prediction of per-sheet processing cost.
    • Collaborated with an IIT Delhi professor on research for layouting answer text types and mapping them to their questions.
    • Handled on-ground operations at CCS University and scaled model deployment to process 500K sheets in 3 weeks.

  2. Visiting Scholar, Medical Imaging · Indian Institute of Technology, Ropar

    Jan 2025 — Jun 2025 · Punjab, India
    • Airway segmentation for diagnosis: segmented airway structures to detect conditions such as adenoid hypertrophy, a key indicator of upper-airway obstruction.
    • Morphological and volumetric analysis: assessed upper-airway shape and volume to evaluate obstruction severity.
    • Currently working on keypoint detection in lateral cephalometric (Lat Ceph) radiographs, experimenting with new methods.

  3. SRIP Intern · IITRAM

    May 2024 — Jul 2024 · Ahmedabad, India
    • Conducted research on PPG signals to develop deep learning methods for monitoring sleep disorders.
    • Created a CNN-based model for classifying sleep disorders using PPG signals, achieving 95.73% accuracy across 6 sleep disorders.
    • Research published in Engineering Applications of Artificial Intelligence (EAAI) journal.

B.E. in Information Technology · UIET, Panjab University · CGPA: 8.52

2021 — 2025 · Chandigarh, India
Projects

Things I've built

Research OS

SvelteKit Supabase Cloudflare Workers AI
  • Built an AI-assisted research paper reader with problem-based organization and highlight-to-note capture.
  • Added a spaced-repetition review system and an AI daily loop for reading prep and memory consolidation.

Low-Resource Punjabi ASR Across Age Groups

Python Wav2Vec2 ASR Speech Processing
  • Trained wav2vec2-based ASR models across age cohorts for low-resource Punjabi speech recognition, achieving a Word Error Rate of 0.17.
Research

Publications & achievements

Publications

Achievements