SYSTEM_ONLINE
NEURAL_INTERFACE.exe

root@cyberspace ~ $ whoami

{'{'} "name": "Prakhar Kothari", {'}'}

root@cyberspace ~ $ cat role.txt

root@cyberspace ~ $ echo $EXPERTISE

AI Research NLP Fine-tuning Computer Vision

0+ PROJECTS
0+ HACKATHON WINS
0 (in progress) Research Papers
Prakhar Kothari
NEURAL_LINK
100%
AI ML NLP
SCROLL

[02] ABOUT_PROFILE.exe

BIO.md
ACTIVE

AI/ML Engineer & Researcher

Currently pursuing B.Tech in Computer Science at the Indian Institute of Information Technology, Nagpur, specializing in Artificial Intelligence and Machine Learning with a CGPA of 8.31/10.

My expertise lies at the intersection of applied AI and research, with a focus on Natural Language Processing, Computer Vision, and Large Language Models. I'm particularly passionate about Retrieval-Augmented Generation (RAG) systems, LLM fine-tuning, and building intelligent applications.

Currently working as a Generative AI Intern at Tecnod8.ai, where I've engineered custom RAG pipelines achieving 95% improvement in information retrieval accuracy.

Research Focus

NLP • RAG Systems • LLM Fine-tuning • Computer Vision

Tech Arsenal

PyTorch • TensorFlow • LangChain • Transformers

Recognition

Winner IIT Roorkee • 2nd NPCI • 3rd IIM Indore

Current Mission

GenAI Intern @ Tecnod8.ai Building RAG Pipelines

[03] WORK_HISTORY.log

OCT_2025 → PRESENT

Generative AI Intern

Tecnod8.ai // REMOTE

  • Engineered custom RAG pipeline achieving 0.40 → 0.78 NDCG improvement (95% boost)
  • Fine-tuned transformer-based embeddings for semantic similarity optimization across domain-specific datasets
  • Optimized GPU utilization on Azure VMs with parallel processing and mixed-precision training
RAG Transformers Azure PyTorch Embeddings

[04] PROJECTS.db

01

Engage-GPT

Fine-tuned Mistral-7B using QLoRA on 1,000+ YouTube pairs. Achieved 70% loss reduction training only 0.79% parameters.

70% Loss Reduction
0.79% Params Trained
Mistral-7B QLoRA GPTQ PyTorch
VIEW_PROJECT
02

Intelligent Research Aggregator

Multi-source platform using LangGraph for parallel aggregation. 70% time saved through AI synthesis.

70% Time Saved
3 Data Sources
LangGraph LangChain Ollama APIs
VIEW_PROJECT
03

upcharAI

Healthcare platform with MobileNet disease prediction. 90.24% accuracy with RAG medical assistant.

90.24% Accuracy
30 Diseases
MERN MobileNet LLaMA Neo4j
VIEW_PROJECT
04

Multimodal RAG

Advanced PDF processing with text and image analysis using multimodal embeddings.

RAG LangChain Multimodal Vector DB
VIEW_PROJECT
05

Image Captioning

Deep learning caption generation with CNN-RNN architecture.

Computer Vision CNN RNN TensorFlow
VIEW_PROJECT
06

PDF QA with AstraDB

Document Q&A system with LangChain & Groq and vector similarity search.

LangChain Groq AstraDB Vector Search
VIEW_PROJECT

[05] TECH_STACK.json

Languages

Python Java C/C++ Kotlin SQL JavaScript

AI & ML

PyTorch TensorFlow Keras Scikit-learn Hugging Face Transformers OpenCV

Gen AI & LLMs

LangChain LangGraph Ollama CrewAI RAG Fine-tuning LoRA QLoRA

Backend & DB

MySQL MongoDB Neo4j Vector DBs FastAPI Flask

[06] ACHIEVEMENTS.sys

01 ST

WINNER

Case-Based Hackathon

IIT Roorkee

2025
02 ND

RUNNER-UP

AI Hackathon by NPCI

IIT Roorkee

2025
03 RD

THIRD PLACE

National Hackathon

IIM Indore

2025
04 TH

FOURTH PLACE

National Robotics

IIT Bombay (e-Yantra)

2020
05 TH

FINALIST (TOP 5)

HACKATRON

IIIT GWALIOR

2024

[07] CONTACT.init

Let's Connect

Open to discussing new projects, research opportunities, or collaborations in AI/ML. Reach out through any channel below.

MESSAGE.send()