Ajeet Kumar
I am currently looking for Job, was a Research Intern and Project Assistant at the Cloud Computing Lab and HIPC Lab, IIT Delhi under Prof. Abhilash Jindal, where i had focused on building and evaluating LLM-based tools called Speculator and design multi-agentic systems Lean4Agent
Before joining IIT Delhi, I completed my M.Sc in Mathematics and Computing at Banaras Hindu University (BHU), Varanasi, where I worked at the DST-CIMS. My Master’s thesis was centered around Discrete Differential Geometry and its Applications, supervised by Prof. Bankteshwar Tiwari.
My academic journey began with a B.Sc (Hons) in Applied Mathematics from Jamia Millia Islamia, New Delhi. Along the way, I also pursued a Data Science Specialization through NPTEL (IIT Madras), strengthening my skills in Programming, Data Analytics, Machine Learning, and Large Language Models.
Research Experience
Speculator - LLM System for OpenAPI Specification Generation from Source Code.
Lean4Agent - Multi-Agent System for Lean4 Code and Proof Generation from Mathematical Problems(Natural Language).
- Solving Wave PDE by HHL Quantum Algorithm(Qiskit)
- Implemented and Practiced Variational Quantum Algorithms like VQEs.
Machine Learning Intern developing end-to-end ML pipelines
Latest Posts
Latest Projects
Lean4Agent is a collaborative AI system designed to tackle complex mathematical problems using multiple intelligent agents.
Lean4Agent: Multi-Agent AI System for Verifiable Writing and Proving Theorems in Formal Mathemtics
Lean4Agent is a collaborative AI system designed to tackle complex mathematical problems using multiple intelligent agents.
NextQuestAI is a high-performance, multi-agent research orchestrator powered by LangGraph and NVIDIA NIM.
NextQuestAI: Deep Research Multi-Agent System
NextQuestAI is a high-performance, multi-agent research orchestrator powered by LangGraph and NVIDIA NIM.
A comprehensive, production-ready Discrete Exterior Calculus (DEC) / FEM implementation for triangle meshes with spectral geometry, curvature analysis, PDE solvers, and advanced vector field processing.
DDG — Advanced Discrete Differential Geometry Framework (Research-Grade)
A comprehensive, production-ready Discrete Exterior Calculus (DEC) / FEM implementation for triangle meshes with spectral geometry, curvature analysis, PDE solvers, and advanced vector field processing.
Autonomous machine learning research framework for AI agents
AutoResearch: Autonomous ML Research Framework
Autonomous machine learning research framework for AI agents
Mathematical Data Driven Methods for Modelling and Simulations of Science and Engineering Problems, mostly modeled as partial differential equation therefore how to solve them for performing simulation of the system. I have trained ViT model onto the Allen Cahn PDE to dataset and checked the model predicted solution for the unseen input dataset. results are close to exact solution.
Vision Transformer(ViT) from Scratch - Solving Partial Differential Equations(PDE)
Mathematical Data Driven Methods for Modelling and Simulations of Science and Engineering Problems, mostly modeled as partial differential equation therefore how to solve them for performing simulation of the system. I have trained ViT model onto the Allen Cahn PDE to dataset and checked the model predicted solution for the unseen input dataset. results are close to exact solution.
How to write and solve computation problem efficiently on different architectural devices having parallel and distributed computational frameworks
Parallel and Distributed Computing
How to write and solve computation problem efficiently on different architectural devices having parallel and distributed computational frameworks
Analyzing sales data from a supermarket MySQL database to uncover insights and predict customer behavior. It uses Python, , and machine learning techniques to preprocess data, perform exploratory analysis, and build predictive models for customer types.r 1
Super-Market Analysis and Prediction using Python and MySQL Database
Analyzing sales data from a supermarket MySQL database to uncover insights and predict customer behavior. It uses Python, , and machine learning techniques to preprocess data, perform exploratory analysis, and build predictive models for customer types.r 1
The Disease Detection Web Application, developed in 2022, is a machine learning-powered platform designed to assist in the early detection of diseases. 
Disease Detection Web Application
The Disease Detection Web Application, developed in 2022, is a machine learning-powered platform designed to assist in the early detection of diseases.