a.k
/
back
03

my projects.

A collection of my work and papers which I have worked on.

2 research
·
4 projects
arXivblog post
research

METIS: Mentoring Engine for Thoughtful Inquiry & Solutions

Abhinav Rajeev Kumar, Dhruv Trehan, Paras Chopra
Accepted into AAAI 2026 (AI For Scientific Research Workshop)
IEEE XPlore
research

Analyzing Quantization in Multilingual Transformer Models for Indic Translation

Praneeth Kasiraju; Abhinav Rajeev Kumar, Ramprasath M, Manikandan N
Accepted into IEEE CMSS 2025
project

Joker-LLM

Implemented PlanSearch for generating "funny" jokes by using multi-dimensional evaluations and bias mitigations via LLM-as-a-judge methodology.

project

RAO: Recursive Agent Orchestrator

An agentic system that spawns and manages LLM-driven agents via a master-delegator model in the form of layers. Built for complex multi-agent workflows.

project

ReportEase

An automated legal insight platform using Mistral LLM fine-tuned on BNS & BNSS datasets to analyze FIRs and generate insightful reports.

3rd Place - Rajasthan Police Hackathon
project

Surge Protector

An AI-driven drone and dashboard system that detects stampedes by marking heads in crowds using a convolutional neural network.

Best AI/ML Track - Stanford TreeHacks