Hi! I'm

Akshay Kulkarni

Machine Learning | Software | Data Scientist

Get in touch!

Resume

Or scroll down to find out more about me! 👇🏼

About Me

Image

Data Scientist.

Hi, I'm Akshay, I'm a Data scientist currently based in San Diego, CA. My current work lies in the Computer Vision domain with specific focus on 3D Deep Learning based Image Segmentation. My interests converge at the intersection of DS, SWE & ML. I am a dedicated individual who enjoys solving complex problems, building things and having fun while at it.

Connect with me —

Experience

NCMIR - UCSD



San Diego, CA

July 2020 - Present

Research Data Scientist

NCMIR Lab
  • Developing ad-hoc solutions for dense semantic segmentation & analysis of high-throughput 2D & 3D complex imaging data as well as implementing tooling and methods for image processing and reconstruction in Python
  • Deploying new features and conducting performance evaluations for trained models & algorithms in CDeep3M2, a Deep Learning pipeline to perform segmentation tasks on various cellular ultrastructures
  • Building & productionalizing backend processing micro-services for CIL-Preview, an online tool allowing end-users to rapidly test performance of any pre-trained models shared & hosted on the CIL-CDeep3M Model Zoo

Northeastern   University


Khoury College of
Computer Science

Boston, MA

Jan 2020 - July 2020

ML Research Assistant

Vitek Lab – [MSstatsQC-ML]
  • Collaborating with an international group of researchers to explore and employ tree-based (Ensembles, XGBoost etc.) and Neural network machine learning models to detect sub-optimal operation and enhance longitudinal quality control in mass spectrometry-based proteomics.
  • Implementing efficient parallelized models for the framework included in the open-source distribution of the tool as an R/Bioconductor package.
  • Developed tools for prediction & inference of deviations in instrument performance using statistical analysis, data simulation & machine learning to allow instrument calibration & corrections of runs, saving time & costs.
  • Implemented a web user-interface for on-the-fly data analysis, visualization and exporting in order to improve demoing & usability as well as containerized the application for dev, test and prod environments

Northeastern   University


Khoury College of
Computer Science

Boston, MA

Jan 2020 - Apr 2020

Research Assistant

VR Lab
  • Processing and Cleaning EEG signal data from subjects in VR settings
  • Employing various Machine Learning algorithms to model the relationship between brainwaves and motion sickness in VR to predict if the subject is cybersick.

Northeastern   University


Khoury College of
Computer Science

Boston, MA

Sep 2019 - Dec 2019
Jan 2019 – May 2019

Graduate Teaching Assistant

- CS3200-[Database Design]
- DS4100-[Data Collection, Integration and Analysis]
  • Assisting the professors by tutoring students during recitations/Labs
  • Designing coursework and assignments, evaluating student projects and providing learning support to students.
  • Guiding and teaching students individually or in groups to help them master assignments and to reinforce learning concepts such as tidying ,storing and analysing data, employing modelling techniques regression, clustering and other machine learning techniques in R.

PREDIKLY


Pune, IN
Feb 2017 - May 2018

Junior Analyst

  • Performed data cleaning, generation of summary statistics, visualizations and contributed in development of dashboards & predictive analytics solutions with client sourced data.

  • Worked as part of the sentiment analysis team, assisting in mining and tokenization for large corpora of texts ranging from customer support incidents, product reviews, social media feeds, news articles, documents, and other sources.

ZIO
TECHNOLOGIES


Dubai, U.A.E.
Aug 2015 – Jan 2016

Sales and Statistics Intern

  • Acquisitioned, managed and analysed product statistics/data for projects and tender submissions and assisted on Systems and AV integration for large scale projects in Dubai for the Government of Dubai and other major sites.

INFRATECH CONTROLS L.L.C


Dubai, U.A.E.
May 2014 – Aug 2014

Projects Research Intern

  • Sales and on-site assistance on projects in B.M.S. (Building management systems) Installations in various business and healthcare sites in Dubai.

Skills

Projects



Image

Neural and Classical Learning to Rank on MSLR-Web10K

Exploring query/document ranking in search engines with tree-based models like LamdaMART using LightGBM and Neural approaches like DNNs using TF-Ranking.

—

Image

Community Detection on Complex Twitter Networks

An implementation of Graph based community detection algorithms with Python, JavaScript and Vue to visualize clusters of Hashtag Networks.

—

Image

Distributed Low-Rank Matrix Factorization Using ALS

An implementation of distributed Alternating Least Squares algorithm to perform matrix factorization in Scala with Apache Spark.

—


Image

Classification of chest radiographs using Convolutional Neural Networks

Classification of chest x-rays from the NIH Chest X-ray Dataset containing 112,120 patient recordsinto one of the 14 possible conditions (Cardiomegaly, Effusion, Pneumothorax etc.)

—



Image

Decision Support Framework/ Multi Objective Optimization Tool

A multi objective optimisation tool to aid user in comparing multiple optimal trade-offs between objectives- Deployed on Heroku

—



Image

Gaussian process classification for O*NET Occupations

Classification task using Gaussian Process algorithm to classify occupations from the O*NET dataset.

—



Other stuff that I've worked on


Image

Q* Learning for OpenAI Taxi-v3

—

Trained an RL agent using the Q-Learning Algorithm in the OpenAI TaxiV3 environment.

Image

Search Engine from Scratch in Python with Lucene

—

Information retrieval model in Python and Apache Lucene, Implementing several ranking algorithms such as BM25, QLM & Vector Space Model to rank parsed and cleaned documents.

Image

Gradient Descent with Line Search

—

Batch gradient descent algorithm implementation with Line search to support dynamic learning rate scaling Note : Implementation incorporated into LinearRegession achieved slightly better error than Scikit-learn version on the same dataset.

Image

Social Triangle counting in Twitter Network Graph

—

Used Equi-joins in MapReduce on a subset of a large twitter graph with 80 million+ egdes to count the number of social triangles i.e. A->B->C->A (a 3-clique).

Image

Analyzing Human Development in India

—

Exploratory Data Analysis and Modelling Project in R to analyse Agricultural yield and Women's employment in rural India conducted under the supervision of Prof. Kylie Bemis at Northeastern.


AQI visualization

—

A visualization tableau workbook to display statistics on Air Quality Index data in the united states from 2001-2016.










Blog Posts & Reports

ONET Occupations Analysis Report
This report dives into the task of classifying a given occupation from the the ONET database by implementing various models to estimate a probability of an occupation being Analytical in nature.
Document retrieval engine analysis
Analysis exploring the development and evaluation of several search engines for a provided test collection of scholarly journal abstracts.