About

Hello World

Hi, I am Abhinav Bohra, a graduate from Indian Institute of Technology, Kharagpur with a Dual Degree (Bachelors & Masters) in Computer Science & Engineering and a Micro Specialization in Artificial Intelligence. I like to apply my technical skills, accrued from my engineering studies, to conflate the spheres of technology and innovation. I am passionate about applying AI to new domains and developing products that have a meaningful impact on people's lives.

Whether it is academics or management, challenging real-life problem-solving always attracts me. In the quest for expanding my horizons, I have constantly aimed at pursuing diverse interests. I'm a team player and love to work in collaboration. I also believe in helping the community and like to be a leader.

When not coding, you will most often find me making EDM or supporting Manchester United. GGMU :)

Machine LearningDeep LearningNatural Language ProcessingComputer VisionSoftware Engineering

Experience

Places I have worked and learned

A few internships and research roles where I contributed to applied ML, process mining, video understanding, and software tools.

Applied Scientist Intern

Amazon · Advertising Technology

Bangalore

Jan 2022 - Jul 2022

Traffic Quality for Sponsored Ads

Building ML & DL algorithms for detecting advanced robotic behaviors to safeguard Amazon advertisers in an ever-evolving fraud landscape
Developed DNN based crawler bot detection model with 11.9% improved robotic coverage and 4.9% higher Impression IVR

PythonSparkPySparkTensorflow

Research Intern

IBM Research AI · IBM AI for Business Process Automation

Bangalore

Dec 2021 - May 2022

Assessing the Impact of Event Log Quality on Predictive Process Monitoring

Carried out statistical analysis of data quality issues as a collection of patterns to improve performance of predictive process models

PythonPyTorchRPM4PYTime Series

Research Intern

Adobe Inc. · Big Data Experience Lab

Bangalore

May 2021 - Aug 2021

Videos2Doc: Generating Documents from a Collection of Procedural Videos

Developed a machine learning-based framework for automated document generation from a collection of procedural videos
Eased the task of making multimedia accessible through automation in conversion to alternate presentation modes
Established the utility of this work via empirical and qualitative evaluations
Published at ACM IUI 2022 — https://dl.acm.org/doi/abs/10.1145/3490100.3516460

PythonPyTorchOpenCVTransformers

Software Development Intern

Philips India Ltd. · Philips Innovation Campus

Bangalore

Apr 2020 - Jun 2020

Application of AI in prediction of Software Vulnerabilities

Designed, developed & deployed ML-based software to predict software vulnerabilities for optimizing the software testing process
Fine-tuned SVM, Multinomial Naive Bayes, and Dense Neural Networks (LSTMs) for multiclass classification with 30+ categories
Developed user interface using HTML5, CSS & JavaScript, integrated with backend using Eel

PythonPython-EelKerasNLTKPandasScikit-Learn

Research Intern

Indian Institute of Management, Ahmedabad · Department of Production and Quantitative Methods

Ahmedabad

May 2019 - Jul 2019

Application of Machine Learning in Finance

Implemented an unsupervised, graph-based topic modeling algorithm to extract topic-specific articles from the news feed
Successfully extracted 1,000+ articles from the CNN DailyMail dataset

PythonNumpyNLTKPandasNetworkx

Open Source

Research and software projects

Research projects

Research projects from the pre-ChatGPT era: when NLP actually had 11 distinct tasks and you couldn't just prompt your way to state-of-the-art

ECTSum: Bullet Point Summarization of Long Earnings Call Transcripts

EMNLP 2022 (Main Conference)

Created ECTSum dataset using Earnings Call Transcripts of publicly traded companies. Benchmarked with BigBird, SummaRuNNer, Longformer. Proposed FinBERT-T5 paraphraser with 13.3% ROUGE-2 gain and 8.5% less factual hallucination.

PythonPyTorchCUDA

Video Game Level Generation using DCGAN

Aug 2022 - Nov 2022

Explored GAN architecture for generating playable Super Mario game levels. Compared latent space search techniques to optimise GAN inputs.

PythonJavaBash

Neural File Search Engine

Aug 2022 - Nov 2022

Designed CoeuSearch, an NLP-based local file search engine using SBERT dual encoders and KeyBERT Topic Extraction. Three-fold search strategy with cache optimization reducing response time by 70%.

PythonPyTorchNLTKDjango

Multilingual News Article Similarity

Jan 2022 - Apr 2022

Used mBERT and XLM for article similarity prediction. Sentence Transformer + cosine similarity approach. Ranked 19th in SemEval 2022 Task 8 Leaderboard with Pearson correlation 0.721.

PythonPyTorchCUDA

Entailment as Few Shot Learner For ACOS Quad Extraction Task

Dec 2021 - Apr 2022

Two novel approaches for ACOS task: entailment-based category classification and paraphrase modeling paradigm. Reported new state-of-the-art results in both in-domain and cross-domain settings.

PythonPyTorchPaddle-NLPCUDA

Investigating Generative Approaches For ACOS Quad Extraction Task

Aug 2021 - Nov 2021

Three generative methods for Aspect Category Opinion Sentiment (ACOS): two autoregressive and one non-autoregressive with bipartite matching loss. State-of-the-art results on benchmark dataset.

PythonPyTorchFast-AICUDA

Multitasking Framework for Emotional Analysis

Jan 2021 - Apr 2021

Implementation of multi-task ensemble framework combining CNN, LSTM, GRU and hand-crafted features for emotion, sentiment, and intensity prediction. 5.2% accuracy increase and +0.33 Pearson correlation on emotion tasks.

PythonKerasTensorflow

Stock Price Movement Prediction using Sentiment Analysis

Aug 2021 - Nov 2021

BERT sentiment analysis on company tweets combined with LSTM stock price prediction using OHLCV features. Statistical correlation between social media sentiment and stock movement.

PythonKerasTime SeriesSequence Models

Software builds

Software builds before 'vibe-coding' was a thing and Stack Overflow threads were sacred texts you bookmarked

Social eCommerce WebApp

HAVE, Hybrid Automated Virtual Enterprise

Jan 2021 - Apr 2021

Social eCommerce platform with 'Team Purchase' concept where bulk buying reduces per-unit cost. Users form groups to share reduced prices.

PythonPython-DjangoMySQLHTML5CSS3

Peer-2-Peer Chat Application

Jan 2021 - Apr 2021

P2P chat application with no central server. Uses TCP as transport protocol. Every instance runs a peer-chat server listening for connections.

CBash

THAT Hearing Assistance & Transcription

Sept 2020 - Nov 2020

Online learning platform for inclusive education. Provides live transcription/subtitles during classes and speech pace assistance for hearing-impaired users.

PythonPython-FlaskJavaScriptMySQLiteHTML5CSS

GrowFast, Employee Management System

Jan 2020 - May 2020

Integrated management tool and Employee-HR portal for top-level management to handle HR activities and internal tasks.

JavaJavaFXMySQL

Other projects

PrivML

Implementation of privacy-preserving SVM with partial homomorphic encryption

Adversarial ML

Generating Adversarial Samples for Attacks on Image Classification Systems

Big Data Processing

Exploring and Implementing Scalable Data Processing Techniques

MLFA

Real-World Machine Learning: Programming Assignments for AI42001

Decision Trees

Implementation of Decision Trees using ID3 algo without any off-the-shelf library

Naive Bayes

Implementation of Gaussian & Multinomial Bayes Classifier without any off-the-shelf library

G-audio Plugin

A plugin for automatic extraction of user's Google dashboard data

PintOS

A simple operating system framework for the 80x86 architecture

tinyC Compiler

A compiler for a C-based programming language

Command Line Interpreter

A rudimentary command-line running on Linux OS

Socket Programming

Networks Lab Assignments on Socket Programming

Network Namespaces

Network Namespaces, Ethernet Bridge & Routes

Producer Consumer System

A multiple producer-consumer system for creating prioritized jobs

Graham Scan

Algorithm for finding the convex hull of a finite set of points in the plane in O(N*log(N))

Ford Fulkerson

A method for computing maximum flow in a network using Edmonds-Karp algorithm

Packet Sniffer

Packet Sniffer & Packet Analyzer using Wireshark

Teaching experience

Classroom, lab, and peer mentoring roles across machine learning, compilers, and student development.

Teaching Assistant

Machine Learning, CS60050

Jan 2023 - Apr 2023

•Facilitated discussions and provided clarifications on complex machine learning concepts to students
•Designed ML assignments and provided constructive feedback on problem-solving, algorithm selection, and code optimization

Teaching Assistant

Compilers Lab, CS31003

Aug 2022 - Nov 2022

•Coordinated lab sessions for a class of 120 students, overseeing the setup, assessment, and resolution of queries related to assignments

Student Mentor

Student Welfare Group, IIT KGP

Jul 2020 - Jun 2022

•Mentor for 5 junior batch students, offering guidance on academic and personal matters as their first point of contact

Subhead

Business Club, IIT KGP

Jul 2019 - Jun 2020

•Speaker at biweekly knowledge sessions on machine learning algorithms such as Regression, Clustering, etc.

Skills

Tools I have worked with

Programming languages, libraries, and frameworks I have dabbled with.

Achievements

Secured Department Change to 'Computer Science & Engineering' with Institute Rank 12 among 1465 students (Top 1%) in first year
Recipient of Sunil Kanti Dant Memorial Award for excellence in 'English for Communication'
Achieved Global Rank 12 in CodeChef Starters Challenge Sept 2021
Achieved Rank 148/20,785 in Sept Long Challenge 2021 (Username: abohra)
Silver Medalist in Facebook Hack Coding Challenge 2021