Open to AI/ML and data roles

Amteshwar Singh

I build generative-AI applications and multi-agent systems, and turn messy datasets into decisions people can act on.

About

Where I'm coming from and what I'm working on.

I'm a fresher with a strong interest in artificial intelligence, machine learning and data analytics. My work spans the full data lifecycle - from cleaning and visualising complex datasets to building scalable applications on top of generative AI.

I focus on intelligent systems built with frameworks like LangChain and CrewAI, backed by analysis in SQL, Tableau and Power BI. The part I enjoy most is translating raw data into insights people actually use.

Right now I'm exploring agentic AI - orchestrating workflows with CrewAI and the Model Context Protocol to build systems that can run on their own.

Technical skills

Tools and frameworks I've built with.

Languages & frameworks

  • Python
  • SQL
  • FastAPI
  • Flask
  • Streamlit

Generative AI & LLMs

  • RAG
  • Fine-tuning (LoRA / QLoRA)
  • OpenAI
  • Hugging Face
  • Ollama
  • Groq
  • NVIDIA NIM
  • Google Gemma
  • Meta Llama

Agent frameworks

  • LangChain
  • LangGraph
  • LangSmith
  • CrewAI
  • Model Context Protocol

Machine learning & NLP

  • TensorFlow
  • Scikit-learn
  • NLTK
  • Text embeddings
  • Feature engineering
  • Predictive modeling

Data & analytics

  • Pandas
  • NumPy
  • Matplotlib
  • Seaborn
  • EDA
  • Power BI
  • Tableau
  • Excel

Databases & platforms

  • FAISS
  • Chroma
  • Pinecone
  • Neo4j
  • AstraDB
  • Snowflake
  • Git & GitHub
  • Docker
  • AWS (basic)

Featured projects

Three things I've shipped recently.

Autonomous marketing agent

A multi-agent system built with CrewAI that runs real-time market research and produces a full go-to-market plan from a single prompt. Pydantic enforces the output schema, so SEO blogs, social calendars and video scripts come back API-ready.

  • Python
  • CrewAI
  • OpenAI
  • Pydantic

Agentic search engine

A reasoning engine built on LangChain that routes each query to the right source - live web search, arXiv or Wikipedia - based on context. A just-in-time RAG pipeline vectorises any URL you paste for instant Q&A, with Groq handling inference.

  • Python
  • LangChain
  • Groq
  • OpenAI
  • FAISS
  • Streamlit

Loan default risk dashboard

A Power BI dashboard analysing loan distribution, borrower demographics and default patterns. It surfaces high-risk segments by credit score, income and employment type so lenders can see where their exposure actually sits.

  • Power BI
  • DAX
  • Power Query
  • Data modeling