Gunjan_

/ / ai/ml developer | generative ai | agentic ai

Pawan Kumar Gunjan

AI/ML Developer with 3+ years of hands-on experience in Generative AI, Applied NLP, and Computer Vision. I specialize in Transformers, Multimodal LLMs, Vector Databases, and Agentic AI systems, with a strong focus on turning research ideas into production-ready solutions.

Explore Projects More About Me
3+ Years in AI/ML
2+ Live functional projects
Core Agentic AI, NLP, Computer Vision

[01] / Current focus

Specialization: Generative AI & Agentic AI
Core domains: NLP, Computer Vision, RAG
Current role: AI/ML Computational Science Sr. Analyst

[02] / Core tooling

  • Python
  • LangChain
  • LangGraph
  • Hugging Face
  • FastAPI
  • Docker
  • FAISS
  • ChromaDB
Featured work #portfolio #ai-ml

Building AI systems from research to practical deployment

Selected highlights include an AI stock research agent, intelligent voice bots, OCR pipelines, machine translation workflows, and production-ready ANPR systems built with modern MLOps practices.

Updated Sep 2026 Open projects →

Selected work

#agentic-ai

Inventment Agent

Built an AI stock research agent that maps natural-language queries to NSE-listed equities, retrieves live quotes and news, and returns structured JSON plus a CLI-friendly report.

  • LangGraph StateGraph orchestration with state checkpointing
  • DuckDuckGo web/news retrieval and structured result ranking
  • Entity extraction and deterministic NSE symbol resolution with ambiguity handling

#computer-vision

Automatic Number Plate Recognition

Developed and deployed ANPR workflows covering data preparation, model training, inference pipelines, and production rollout for efficient license plate recognition.

  • YOLOv8-based plate detection
  • Fine-tuned TrOCR for high-accuracy character recognition
  • Achieved 98% recognition accuracy in real-world conditions

#voice-ai

Intelligent Voice Bot

Designed a conversational voice workflow that converts audio input to text, provides conversation history to an LLM, and returns spoken output using text-to-speech.

  • Speech recognition and whisper-based transcription
  • Context-aware prompting with previous conversation history
  • Audio classification for interruption and exit phrases

#translation

Machine Translation App

Fine-tuned Helsinki-NLP Opus-MT models for Hindi-English and English-Hindi translation and packaged the capability into an accessible translation application.

  • Model fine-tuning for improved translation quality
  • Application design for bilingual translation workflows
  • End-to-end pipeline from model tuning to user-facing app

Experience

03/2026 – Present

Noida, India

AI/ML Computational Science Sr. Analyst, Accenture

  • Focused on practical AI deployment and scalable solution design.
  • Worked on research-to-production workflows for modern AI systems.

08/2024 – 02/2026

Delhi, India

Software Developer - AI/ML, Metro Infrasys

  • Built and deployed ANPR systems across data, model, and inference stages.
  • Developed FastAPI services and Docker-based deployment workflows.

12/2022 – 07/2024

Noida, India

TCE (AI & Machine Learning), GeeksforGeeks

  • Worked across Generative AI, NLP, Deep Learning, and Computer Vision.
  • Built hands-on experience in applied AI research and implementation.

Skill stack

Python R HTML CSS LangGraph CrewAI BeeAI AutoGen LangChain LlamaIndex FAISS ChromaDB Hugging Face OpenAI Gemini Claude Ollama NLTK SpaCy Gensim PyTorch TensorFlow Keras OpenCV YOLO Pandas NumPy Matplotlib MLflow ONNX Docker FastAPI