About Me

Machine Learning Engineer @Samespace | Generative AI, LLM, MLOps, Language Modeling

  • ML Engineer with 2+ years of experience in building and deploying ML, LLMs and Transformer Models.
  • Proven ability to optimize model performance and scale to large datasets.
  • Expertise in cloud computing and MLOps.

Download CV

Areas of Interest

Take a look at some of the things I love working on.

MLOps

I build CI/CD pipelines, Docker Containers and use cloud platforms like GCP and AWS for automated model updation and deployment.

Generative AI

I utilize Gen AI to build products like chatbot and voicebot to address customer queries.

ML/AI

I am passionate about learning new innovations and contributing to the field of ML and AI.

  • Qualifications
  • Certifications

Bachelor of Engineering in Computer Science

University of Mumbai

2018 - 2022

⚡ I completed by bachelors with the CGPA of 9.33/10.

⚡ I have studied basic software engineering subjects like DS, Algorithms, DBMS, OS, CA, AI etc.

⚡ Apart from this, I have done few courses on Deep Learning, Data Science, Machine Learning and Python.

Visit Website
Oct 2022 - Present

Samespace

Machine Learning Engineer

- Built conversational voice and chatbot to resolve 80% of user queries by understanding large-scale support docs and knowledge base using various, word embedding techniques, RAG, prompt tuning of GPT-4 and LLAMA-2

- Engineered and assembled an LLM inference server using vLLM, Ray Server, and optimization algorithms like FlashAttention- V2, and PagedAttention, specifically for the Vicuna-13b and Falcon-13b model

- Streamlined Meeting Minutes Creator with ASR, achieving a 98% transcription accuracy rate and enabling summarization and personalised to-do list generation

- Optimized the existing search engine’s performance by 60% by implementing semantic search and image search

- Constructed a productivity tracking system using machine learning techniques, resulting in a 15% productivity increase

Sep 2021 - Aug 2022

Enterpret

Machine Learning Intern

ML Modelling and Packaging
  • Developed a text-classification model based on the business objectives and KPIs, which is capable of predicting the sentiment of given feedback with an F1 score of 84%.
  • Conducted various inter-annotator agreements with the annotation team and fine-tuned the annotation guidelines to minimize potential errors in human-led labeling tasks.
  • Worked on the text-generation problem and created a model by leveraging GPT3, which is capable of predicting a short one-line summary of the given feedback to provide better insights into customer feedback data.
  • Worked on the problem of NLI and was able to increase the performance of the existing model by 30% using the BERT architecture, and hyperparameters tuning.
  • Reduced the inferencing time of ML services by 5% by converting them to ONXX.
ML Lifecycle Management
  • Designed and implemented a drift detection system to monitor the degradation of the predictive quality of the ML models by measuring statistical bias in prediction.
  • Developed various test strategies to make systems more reliable and fault-tolerant.
  • Optimized CI/CD pipelines by reducing the build time of the ML system by 37%.
  • Created an anomaly detection system, which is in charge of raising an alert and reflecting potential anomalies on an observational platform based on some alerting rules.
MLOps
  • Designed and Deployed container-based dynamic deployment pipelines for inferencing by implementing serverless functions along with necessary API gateway handlers.
  • Worked on optimizing the containerizing strategies by leveraging multistage build that reduced the build cost by 7%.
Enterpret Hackathon
  • DeepKey - Developed a keyword extraction system that automatically extracts the most important words from a text using BERT, N-grams, and cosine similarity.
  • Spam Detection System - Trained a spam detection model which makes use of MiniLMv2 and LongFormer architectures with a performance score of 85%.
View Certificate
Jul 2020 - Sep 2020

Algoritmo Lab

Data Science Intern

- Performed statistical analysis on 100,000+ transaction data for threshold tuning to flag alerts in monetary transactions.

- Optimized CAPEX spent on anomaly detection system by 60%.

View Certificate

PdfChat: Chat with Any PDF

PDFChat is developed by leveraging LLM to answer any question about the given pdf.

Abstractive Text Summarizer

Generate grammatically correct and insightful summaries for a given article.

Data Science Portfolio

Data science projects completed by me for academic, and self-learning.

Analyze Diaster Tweets

Built a model capable to filter disaster tweet using NLP.

Automatic Door Unlock System

ADUS uses real-time facial recognition to detect a person's face.

Bigfoot

Developed a platform where one can learn and implement data science in a friendly way.

Stock Screener

Screens five years of historical data of any stock and stores it's in respective CSV format.

Encrypted Data Transfer

Wireless Encrypted transfer of transfer text, files and images between two devices.

Disk Scheduling Algorithms

Implemented various disk scheduling algorithms in c.
More Projects

Programming Languages

Python 95%

Java 70%

C 75%

MLOps

Versioning and Configurations
Frameworks
Resource Management
Optimization and Monitoring
AWS
NLP

Software Engineering

Front-End
Deployment
Databases
SE and Compute Tools

Generative AI

LLM
Transformers

A Study of Artificial Neural Networks (ANN)

Dec 7, 2020

Artificial neural networks (ANNs), usually called neural networks (NNs), are computing systems which are inspired by the biological neural networks that constitute human brains.

Regularization Techniques in Machine Learning

Apr 20, 2021

Regularization helps to overcome overfitting while developing machine learning models. These techniques intend to reduce the risk of overfitting without increasing the bias significantly.

Understanding Logistic Regression

Apr 23, 2021

Logistic regression is a method of classification in which the model outputs the probability of a categorical target variable Y belonging to a certain class.

More Blogs