Harsh Garg

DevOps Engineer

Bengaluru, Karnataka, India2 yrs 2 mos experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

AI enthusiast with experience in large language models.
Developed innovative solutions for knowledge acquisition.
Strong foundation in machine learning and data analytics.

Stackforce AI infers this person is a Machine Learning and AI specialist with a focus on innovative solutions.

Contact

Skills

Core Skills

Machine LearningArtificial Intelligence (ai)

Other Skills

AI validation processAPI DevelopmentAPI GatewayAWS serverlessAmazon Web Services (AWS)AnalyticsApache BeamArtifact RegistryBERT (Language Model)Bayesian Neural NetworkC (Programming Language)C++Cloud ComputingCommunicationConcept Index

About

I am a B.Tech graduate from Indian Institute of Technology Kanpur. I am fond of probability and statistics and working with innovations that can benefit society. I believe that artificial intelligence is one of the technologies that can benefit a large number of people in a variety of ways. I am an AI enthusiast who has worked with large language models, machine and deep learning models, and API development. If you need my experience, reach out to me at any time.

Experience

2 yrs 2 mos

Total Experience

1 yr 1 mo

Average Tenure

1 yr 6 mos

Current Experience

Exl

2 roles

Assistant Manager - Cloud and AI Developer

Oct 2024 – Present · 1 yr 6 mos · Noida, Uttar Pradesh, India

Trainee

Sep 2024 – Oct 2024 · 1 mo · Noida, Uttar Pradesh, India

Hurrey

2 roles

Data Scientist

Jul 2024 – Aug 2024 · 1 mo · Bengaluru, Karnataka, India · On-site

Machine Learning Intern

Dec 2023 – Jul 2024 · 7 mos · Bengaluru, Karnataka, India · On-site

OBJECTIVE
Contributed to the development of an Augmenting Intelligence Computing that enhances knowledge acquisition, validation, and individualized learning by integrating human intelligence with machine learning
STRATEGY
Prepared a 5M row SQL dataset using SQLite, with data extracted and classified through a fine-tuned BERT
model for concept classification and Flair NER for entity recognition
Used Graph Neural Networks and LLaMA model to tailor content based on user understanding, enhancing
engagement and learning outcomes
Employed a bipartite GNN to model user-topic relationships, extract embeddings, and predict links, resulting
in personalized and highly relevant topic recommendations
Implemented an AI validation process involving co-reference resolution, claim extraction, evidence retrieval,
LLM analysis, and to ensure the accuracy and reliability of user contributions
Designed and implemented key metrics for AIC, including Concept Index, Domain Index, Correctness
Score, Reasoning Score, Content Fitment Index, and Language Index, to tailor content based on user
proficiency, domain expertise, accuracy, reasoning abilities, content relevance, and language comprehension

SQLBERT (Language Model)Graph Neural NetworksLLaMAAI validation processConcept Index+7

Byajbook (backed by 100x.vc)

Machine Learning Intern

May 2023 – Jul 2023 · 2 mos · India

Indian institute of technology, kanpur

Undergraduate Student Researcher

Feb 2023 – Oct 2023 · 8 mos · Kanpur, Uttar Pradesh, India · On-site

Mentors: Prof. Arnab Bhattacharya and Prof. Shivam Tripathi
Objective:
Classify fog in optical camera images using visibility data, advanced classification, and optimal features
Strategy:
Employed sensor-gathered visibility values to classify optical camera images as fog and no-fog images
Categorized fog images into varying density levels: very dense, dense, medium, light, and no-fog
Mitigated class imbalance by visually representing fog and no-fog data across timeframes using bar plots
Utilized VGG16 model for feature extraction from Raw Images, followed by feature selection through PCA
Determined optimal feature count using Scree Plot and Explained Variance Ratio
Executed SVM, Random Forest, Logistic Regression, ANN models for Binary & Multiclass Classification
Impact:
The NN model achieves the highest test accuracy (85.54% binary and 78.25% multiclass) among all models
Bayesian Neural Network used to forecast fog probability, yielding a mean CSI of 0.69 on the test dataset