Skip to content
View sukeshan's full-sized avatar

Block or report sukeshan

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
sukeshan/README.md

Sukesh ๐Ÿ“บ

Senior Research Engineer
Building Intelligent Systems @ Shoppin'๐Ÿ“

NLP Expert Generative AI AWS Certified


๐Ÿ‘จ๐Ÿ’ป About Me

AI Solutions Engineer with 2+ years specializing in production-grade machine learning systems.
Focus on delivering enterprise NLP solutions through:

โœ“ Scalable LLM Architectures      โœ“ Optimized Model Inference
โœ“ Context-Aware RAG Systems       โœ“ Cloud-native Deployments
โœ“ Custom Embedding Spaces         โœ“ High-throughput Data Pipelines

Driven by metrics-focused development and cutting-edge AI research.

๐Ÿš€ Key Achievements

Candidate Pre-Screening Agent

NPS 4.6/5
  • ๐Ÿš€ Served 150K+ candidates through AI chatbot
  • ๐Ÿง  Integrated Quantized Llama-2 (7B) with RAG Flow
  • โ˜๏ธ AWS Lambda deployment with >99.9% uptime
PyTorch VectorDB RAGAS

Recommendation System

300K Candidates
  • โฑ๏ธ 2 minute search time for 300K+ profiles
  • ๐Ÿ’พ 40% memory reduction via BFloat16
  • โšก Flash Attention optimization for embeddings
MongoDB FAISS Custom Embeddings

๐Ÿ”ฌ Active Research Areas

๐Ÿงฎ Model Interpretability Framework

โ–ธ Developing token fairness metrics (0-1 scale) โ–ธ Bias analysis in enterprise LLM systems โ–ธ Multi-dimensional fairness evaluation matrix SHAP LIME Attention Analysis

๐Ÿ“ Semantic Embeddings Research

โ–ธ Reducing positive skewness in encoder outputs โ–ธ Contrastive learning for balanced representations โ–ธ Dimensionality impact on semantic clustering Transformers PCA t-SNE
Research Progress 2024
Model Fairness  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 62%  
Embedding Opt   โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 55%

Pinned Loading

  1. Satellite-Image-Segmentaion Satellite-Image-Segmentaion Public

    Customly designed U-Net architecture for counting number of buildings in satellite image and area of each building

    Python 3 1

  2. Identifying-the-purpose-of-the-tablet Identifying-the-purpose-of-the-tablet Public

    Identifying the Drug using only Spatial Information of the Tabletย image

    Python 1

  3. Custom-BERT Custom-BERT Public

    Customized the legacy BERT architecture by integrating recent research advancements focused on model performance optimization

    Python

  4. NumberPlate-Detection-Using-Yolov5 NumberPlate-Detection-Using-Yolov5 Public

    The moto of the project is automate the parking system through camera .Using yolo-v5 model detect number plate and retrieve information using AWS texract.

    Python 1

  5. Recipe_Search Recipe_Search Public

    Retrieve food recipes based on the user given queries using Bipartite Graph Structure

    Python

  6. ReACT-Method-Based-E-commerce-Agent ReACT-Method-Based-E-commerce-Agent Public

    This repository contains code that enables searching for products across various platforms (such as Amazon and Walmart) using a custom agentic framework. The tools provided can search for products,โ€ฆ

    Python 1