Meghana Bhange, Developer in Montreal, QC, Canada
Meghana is available for hire
Hire Meghana

Meghana Bhange

Verified Expert  in Engineering

Bio

Meghana is a machine learning engineer with a passion for solving problems in a data-driven manner. She is currently pursuing her Master's Degree with a research focus on privacy-preserving technologies at the Trustworthy Information Systems Lab, ÉTS Montréal. She has experience in natural language processing and has previously published work at SemEval-2020. Meghana is passionate about working on creative projects and always looks for new ways to apply her skills.

Portfolio

UInclude, Inc
Natural Language Processing (NLP), Machine Learning...
TISL Lab at ETS Montreal
Python 3, Artificial Intelligence (AI), Scikit-learn
Freelance Client
Artificial Intelligence (AI), Chatbots, OpenAI GPT-3 API...

Experience

Availability

Part-time

Preferred Environment

Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), AI Design

The most amazing...

...project I've worked on is developing an end-to-end custom recognition service on resource-constrained code-mixed settings with low latency requirements.

Work Experience

AI Engineer

2023 - PRESENT
UInclude, Inc
  • Developed a context-specific biased word matching model utilizing SpaCy and a Rule-based engine to identify biased words in job listings.
  • Created a synonym enricher employing sentence transformer and GPT-3 to discover context-specific synonyms, replacing biased words in job listings with unbiased alternatives.
  • Deployed the models using a FastAPI endpoint on AWS while storing and querying the data through DynamoDB.
Technologies: Natural Language Processing (NLP), Machine Learning, Generative Pre-trained Transformers (GPT), Amazon Web Services (AWS), Deep Learning, Python, TensorFlow, Scikit-learn, Amazon DynamoDB, Amazon S3 (AWS S3), FastAPI, SpaCy

Researcher

2022 - PRESENT
TISL Lab at ETS Montreal
  • Researched privacy-preserving ML and data publishing for a complaint redressal system.
  • Researched model extraction attacks on machine learning systems with counterfactual explanation APIs.
  • Modeled an adversary that can leverage the information provided by counterfactual explanations to build high-fidelity and high-accuracy model extraction attacks.
  • Benchmarked the model performance on the Folktables dataset, with the extracted model gaining fidelity of around 97.6%.
Technologies: Python 3, Artificial Intelligence (AI), Scikit-learn

AI Developer

2023 - 2023
Freelance Client
  • Developed a FastAPl endpoint for a GPT-4 based chat interface tailored for parents and students. Successfully deployed the application on DigitalOcean, ensuring robust performance and scalability.
  • Enhanced the chat endpoints by integrating it with LangChain, incorporating plugins like Wikipedia, Search, and Math. This integration improved the reliability of information.
  • Utilized a vector database to query documents for reliability of information retrieval. Developed endpoints capable of analyzing chat history to extract relevant topics and concepts.
Technologies: Artificial Intelligence (AI), Chatbots, OpenAI GPT-3 API, Generative Pre-trained Transformers (GPT), ChatGPT, LangChain

OpenAI Developer

2023 - 2023
Zurney.app
  • Built a FastAPI back end with GPT-3 API integration to generate a travel itinerary for a trip and extract locations. These locations were then geo-encoded with co-ordinates.
  • Built a Next.js app to display the travel itinerary and show the geo-locations on Google Maps color-codes corresponding to days in the trip and information about each location.
  • Dockerized and deployed both the FastAPI back end and Next.js front end to DigitalOcean.
Technologies: Artificial Intelligence (AI), ChatGPT, OpenAI, Next.js, FastAPI, DigitalOcean, Large Language Models (LLMs), Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Web Development, OpenAI GPT-3 API

Machine Learning Engineer

2021 - 2022
Hunters.ai
  • Researched and built analytical tools for evaluating threat-hunting detectors and understanding abnormal patterns in detection outputs.
  • Organized the monitoring and quality check infrastructure in machine learning detectors.
  • Created a framework for deep investigation of threats.
Technologies: Python 3, Machine Learning, SQL, Artificial Intelligence (AI), Deep Learning, Scikit-learn, AI Design, Machine Learning Operations (MLOps), Amazon SageMaker, Kubernetes, APIs, Data Pipelines, Amazon Web Services (AWS), PostgreSQL, Data Engineering, Amazon S3 (AWS S3), Amazon DynamoDB

Machine Learning Engineer

2020 - 2021
The Verloop.io
  • Contributed to the intent recognition service using a sentence transformer to improve the top-K recall and accuracy, which improved F1 by 40% absolute.
  • Designed, built, and deployed a multi-lingual name recognition service across all clients.
  • Evaluated the performance of various language models like ULMFiT and VAMPIRE for low-resource language contexts.
  • Created synthetic training data for FAQ systems in a chatbot using Generative Pre-trained Transformer 3 (GPT3) AI.
Technologies: Machine Learning, Python 3, Pandas, SpaCy, Chatbots, OpenAI, Artificial Intelligence (AI), Deep Learning, Scikit-learn, AI Design, Django, Google Cloud, Google Cloud Platform (GCP), Speech Recognition, APIs, Web Development, Text Generation, Language Models, Large Language Models (LLMs), Flask, Data Pipelines, Machine Learning Operations (MLOps), Computational Linguistics, DaVinci, Generative Pre-trained Transformers (GPT), PostgreSQL

Machine Learning Intern

2019 - 2019
The Verloop.io
  • Created a person-name extractor customized for multilingual conversations. Tweaked Flair, Facebook's natural language processing library, to work on low-latency use cases in English, Spanish, and French.
  • Improved the final model achieves by 47% in F1 compared to the previously deployed FastText mode.
  • Deployed the developed multilingual name extractor to production with overall latency of under 500 milliseconds.
Technologies: Artificial Intelligence (AI), Chatbots, Deep Learning, Scikit-learn, Text Generation, Language Models, Large Language Models (LLMs), Flask, PostgreSQL

Model Extraction Attack Using Counterfactual Explanation

A research project that I worked on with the Trustworthy Information Systems Lab at ETS Montreal. We researched how model adversaries can leverage the information provided by counterfactual explanations to build high-fidelity and high-accuracy model extraction attacks.

LitNER | Literature Named Entity Recognition

https://github.com/meghanabhange/litNER
Named entity recognition based on Spacy3 trained on the LitBank dataset. This project uses Roberta XLM as a base model and fine-tuned literature data to understand the terms generally used in literature. The pre-trained model released with the project can also be used to perform NER tasks on any literary text.

Hinglish Twitter Sentiment Detection | SemEval2020

https://arxiv.org/abs/2008.09820
This work adds two common approaches, fine-tuning large transformer models and sample efficient methods like ULMFiT. Prior work demonstrated the efficacy of classical ML methods for polarity detection. We fine-tuned general-purpose languages representation models, such as those of the BERT family, which are benchmarked along with classical machine learning and ensemble methods. We showed that NB-SVM beats RoBERTa by 6.2%. The best-performing model is a majority-vote ensemble which achieves an F1 of 0.707.

Wikipedia Textbook Assistant

https://github.com/meghanabhange/Wikipedia-Textbook-Assistant
Simple streamlit application to be paired with textbooks so that one can easily extract the keyphrases from the text they are reading and get detailed information from Wikipedia to understand the relevant context.

Artificial Insanity (Cards Against Humanity with Stable Diffusion) | Toptal Hackathon

Artificial Insanity is a multiplayer game for the Toptal Hackathon, which uses Stable Diffusion to generate image responses for a text prompt. In each round, a player chooses prompt, and other players generate a response using an integrated Stable Diffusion interface.

I benchmarked performance in terms of quality and latency for DALLE and Stable Diffusion. Also, I deployed the final model on FastAPI to make it easier to integrate with the rest of the back end. The solution won the second prize in the Hackathon.
2023 - 2023

Master's Degree (Ongoing) in Information Technology Engineering

École de Technologie Supérieure - Montreal, Canada

2016 - 2020

Bachelor's Degree in Electronics and Telecommunication Engineering

Savitribai Phule Pune University - Pune, India

Libraries/APIs

Pandas, Scikit-learn, SpaCy, TensorFlow

Tools

Slack, Named-entity Recognition (NER), Amazon SageMaker, ChatGPT

Languages

Python, SQL, Python 3

Frameworks

Django, Flask, Streamlit, Next.js

Storage

Data Pipelines, PostgreSQL, Amazon S3 (AWS S3), Amazon DynamoDB, Google Cloud

Platforms

Kubernetes, Google Cloud Platform (GCP), Amazon Web Services (AWS), Visual Studio Code (VS Code), DigitalOcean, AWS Lambda

Industry Expertise

Cybersecurity

Other

Machine Learning, Natural Language Processing (NLP), Artificial Intelligence (AI), Deep Learning, APIs, Text Generation, Language Models, Data Engineering, Chatbots, OpenAI, AI Design, Machine Learning Operations (MLOps), Large Language Models (LLMs), Computational Linguistics, Generative Pre-trained Transformers (GPT), OpenAI GPT-3 API, Research, Transfer Learning, BERT, Signals, Information Theory, Custom BERT, Stable Diffusion, DALL-E, FastAPI, Inference API, Speech Recognition, Web Development, DaVinci, Systems, Cryptography, Information Technology, Prompt Engineering, LangChain

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring