Param Sharma

AI/ML Researcher · Patent Holder · Builder

Applied AI & RAG Algorithmic Trading System Architecture Maths

Product-focused AI/ML Engineer who ships fast. I turn complex algorithms into scalable products and obsess over the details that make systems robust. I’ve worked on computer vision, generative AI and RAG, algorithmic trading models, and high-performance backends. Mostly with Python, C++, PyTorch, or whatever gets the job done.

Open to Work: Full-Time, Freelance, or Collabs.

Technologies & Tools

Python

C++

Java

TensorFlow

PyTorch

OpenCV

Python

C++

Java

TensorFlow

PyTorch

OpenCV

GitHub Contributions

Featured Projects

Kubernetes MLOps & Observability

End-to-end NLP pipeline featuring DVC data versioning, MLflow tracking, GitHub Actions CI/CD, AWS EKS deployment, and Prometheus/Grafana monitoring.

Read details →

Fasal Saathi AI App

End-to-end crop disease classification pipeline featuring two public datasets (IEEE/Mendeley) and a live Android application deployed on the Play Store.

Read details →

Vehicle Insurance MLOps

Production-grade MLOps pipeline featuring automated model gating via AWS S3, zero-touch CI/CD deployment, and real-time FastAPI inference on EC2.

Read details →

MCQGyaan AI

AI-powered EdTech platform featuring automated generation and context-aware semantic evaluation of handwritten answers using fine-tuned NLP models.

Read details →

Core Research Interests

DEEP LEARNING

Computer Vision & Deep Learning

Building CNN and transformer-based architectures for real-world image classification — including agricultural disease detection from leaf imagery, with a publicly released dataset on IEEE DataPort.

NLP / GEN AI

Natural Language Processing

Exploring advanced natural language processing techniques and Large Language Models (LLMs). Currently researching optimization methods for efficient LLM fine-tuning and deployment.

REINFORCEMENT LEARNING

Reinforcement Learning

Investigating the algorithmic foundations of sequential decision-making, training autonomous AI agents, and applying RL frameworks to optimize complex, dynamic, real-world systems.

MATHEMATICAL ML

Mathematical Optimisations

Exploring the mathematical foundations of ML — loss landscape geometry, non-convex optimization, stochastic processes, and gradient-free search methods for robust model performance.

QUANT FINANCE

Algorithmic Trading & Quant Systems

Data-driven decision modelling for algorithmic and high-frequency trading — statistical signal generation, backtesting frameworks, LSTM architectures, and risk-adjusted performance analysis.

COMPUTER SCIENCE CORE

Backend, System Design & DSA

Developing strong foundations in data structures, algorithms, and backend system design. Focused on building efficient, scalable software and applying algorithmic thinking to computational challenges.

Get In Touch

Let's Connect

Currently open to new opportunities, research collaborations, and open-source contributions. Drop a message!

Email Me Directly

paramsharma802@gmail.com Open Email App

Research Mentors & Supervisors

A dedicated overview of the academic and industry experts guiding my research journey.

ACADEMIC RESEARCH SUPERVISOR

Dr. Jitendra Goyal

Assistant Professor, Dept. of CSE
Birla Institute of Technology, Mesra (Jaipur Extension)

PhD (CSE) from MNIT Jaipur · 8 years teaching experience · Research areas: Blockchain Technology, IoT Security, Information Security & Cyber Forensics · IEEE & IEEE Computer Society Member · PI on Seed Money project on Blockchain-based Academic Credentials

BIT Mesra Profile ORCID

INDUSTRY INTERNSHIP MENTOR

Mr. Vikas Sharma

Founder & Director, MaxBrain Technologies
Gopalpura, Rajasthan, India

Vikas Sharma is an engineering consultant and researcher with over 8 years of expertise in computational modeling. An alumnus of MNIT Jaipur, he specializes in bridging engineering principles with modern AI/ML applications. As a research mentor, he has overseen the development of multiple SOTA projects, including TinyML optimization, hybrid DL architectures, and industrial patents. With a prolific publication record in IEEE and Elsevier, he provides the academic oversight necessary to transition theoretical research into scalable, real-world solutions.

MaxBrain Technologies ORCID Google Scholar

BACHELOR THESIS SUPERVISOR

Dr. Pankaj Gupta

Assistant Professor, Dept. of CSE
Birla Institute of Technology, Mesra (Jaipur Extension)

PhD in Computer Science and Engineering from BIT Mesra with over 19 years of rich academic teaching and research mentorship experience. A prolific researcher with over 20 peer-reviewed publications indexed across Scopus, ESCI, and DBLP, he actively guides doctoral scholars and serves as a technical reviewer for leading international journals, including Bentham Science and IGI Global. His core research expertise lies across Data Science, Machine Learning, and complex Data Mining architectures. He provided the formal academic oversight and core methodology guidance for my Bachelor's Thesis on hybrid transformer-optimizer frameworks.

BIT Mesra Profile

Publications, Datasets & Patents

Preprocessed Indian Ridge Gourd Leaf Image Dataset for Healthy and Diseased Leaf Detection

Param Sharma et al. · IEEE DataPort · Publicly Available Dataset · 9,000 preprocessed images across 3 classes · 2025

IEEE DataPort

TinyML-Based Lightweight Deep Learning Model for Human Activity Recognition on Edge Devices

Param Sharma et al. · WCAIAA 2026 Conference · NFSU Goa · Organised by SCRS · 2026

GitHub Repo

Patent: AI-Assisted Method for Agricultural Leaf Disease Detection and Classification

Param Sharma · Indian Patent · Filed & Granted · 2025

View Patent Document

Conference Paper on Agricultural Leaf Disease Detection — MIND 2025

Param Sharma · MIND 2025 Conference · 2025 · DOI / Proceedings link to be updated (Currently in Press)

Certificate Link

Journal Paper — Discover Artificial Intelligence (Under Review)

Param Sharma (Main Author) · Discover Artificial Intelligence (Springer) · Submitted for Peer Review · 2026

Main author · Currently under peer review. This comprehensive work delivers a systematic and bibliometric analysis mapping the architectural evolution, optimization trends, and deployment landscapes of deep learning models addressing complex agricultural challenges.

Brinjal (Eggplant) Leaf Disease Image Dataset — Mendeley Data

Param Sharma (Main Author) · Mendeley Data · Publicly Available Dataset · 2026

Mendeley Data

Main author of dataset and data in brief paper for this dataset

Data in Brief Paper — Brinjal (Eggplant) Leaf Disease Dataset

Param Sharma (Main Author) · Data in Brief · Elsevier · Under Review / In Press

Main author · Link to be updated upon publication

Mobile Application — Fasal Saathi (Google Play)

Android App · Live on Play Store · 2026

Google Play Store

Main author · Mobile application for crop disease detection and farmer support, integrating machine learning models and real-time data collection.

More publications under review — track on ORCID & Google Scholar

Full citations will be added upon acceptance

ORCID Google Scholar

Recognition & Media

Punjab Kesari Newspaper

Research work featured in Punjab Kesari, one of India's leading Hindi-language newspapers — recognising the contribution to agricultural AI research as an undergraduate student.

View Clipping →

WCAIAA 2026 Conference — NFSU Goa

Research paper on TinyML-based Human Activity Recognition presented at the World Conference on AI Applications and Advancements (WCAIAA 2026).

IEEE DataPort — Public Dataset

Agricultural leaf disease dataset (9,000 images, 3 classes) published and made freely available on IEEE DataPort.

View Dataset →

Projects

Fasal Saathi — AI Crop Disease Classifier & App

An end-to-end deep learning ecosystem for agricultural disease detection. I collected, augmented, and published two major preprocessed datasets—Ridge Gourd (IEEE DataPort) and Brinjal/Eggplant (Mendeley Data). The optimized CNN pipeline is actively deployed in "Fasal Saathi," a live Android application on the Google Play Store helping farmers identify leaf diseases in real-time.

PyTorch CNN Android Play Store App IEEE & Mendeley Datasets Patent Granted

Play Store GitHub IEEE DataPort Mendeley Data

MCQGyaan AI — Smart EdTech Platform

An enterprise-grade educational platform featuring an AI-driven answer evaluation system capable of analyzing both handwritten and typed responses. The core evaluation engine utilizes context-aware NLP models, fine-tuned on a custom-curated dataset to assess semantic meaning rather than simple keyword matching. It features automated question paper generation, real-time performance tracking, and role-based portals for students and educators.

LLM Fine-Tuning NLP Custom Dataset Computer Vision / OCR Full-Stack AI

Visit Platform →

Cloud-Native MLOps & Observability Pipeline

An end-to-end, production-grade MLOps system for IMDb sentiment analysis. Orchestrates reproducible multi-stage DAG workflows via DVC and AWS S3. Features MLflow and DagsHub for experiment tracking, automated CI/CD quality gates via GitHub Actions, and containerized deployment to an AWS EKS (Kubernetes) cluster. Includes full-stack observability with custom Prometheus instrumentation and real-time Grafana dashboards.

AWS EKS (Kubernetes) Prometheus & Grafana MLflow & DagsHub DVC GitHub Actions NLP

GitHub →

End-to-End MLOps Pipeline: Vehicle Insurance

Architected a modular, production-grade MLOps pipeline to predict customer responses to vehicle insurance offers. The system features automated data ingestion from MongoDB Atlas, strict YAML schema validation, and leak-free preprocessing with SMOTEENN. Implemented automated model gating via AWS S3 and a zero-touch CI/CD pipeline using GitHub Actions, Docker, and AWS ECR/EC2 to serve real-time predictions via FastAPI.

MLOps AWS (EC2/S3/ECR) Docker & CI/CD FastAPI MongoDB Scikit-Learn

GitHub →

TinyML-Based Human Activity Recognition on Edge Devices

Lightweight deep learning models (MobileNet1D, CNN+BiLSTM) for real-time HAR on smartphones and microcontrollers. MobileNet1D achieved 92.77% accuracy; quantised to ~1.4 MB with <100ms inference latency and <50mW power on Arduino/ESP32. Research resulted in a paper presented at WCAIAA 2026 (NFSU Goa).

TensorFlow Lite PyTorch TinyML INT8 Quantisation Arduino ESP32 Conference Paper — WCAIAA 2026

GitHub →

Real-Time Deep-Vision Football Analytics System

ACTIVE

Currently building a production-grade computer vision system for real-time football match analysis. The system performs multi-object tracking of players and the ball, computes live possession statistics, generates tactical heat maps, and reconstructs a 2D top-down tactical view from broadcast footage — all in real time using YOLO-based detection pipelines.

YOLOv8 PyTorch OpenCV Multi-Object Tracking Computer Vision Python Ongoing Research

GitHub →