I am passionate about Deep Learning and its impact on all fields encompassing NLP, Quant, and Computational Biology.
Currently, I work as a Quant at JPMorgan Chase.
I aim to make AI interpretable and clinically meaningful!
Developed an unsupervised model to disentangle pose and semantic latent space for cryo-ET subtomograms, making the model pose invariant
Created a dataset spanning 90+ domains for temporal QA on semistructured tabular data, (Wikipedia Infoboxes), the paper was accepted at EMNLP
Conference
Analysed sentiments across five different versions of Sermon on the Mount, using transformers & analysed verse-by-verse polarity scores, to draw historical inferences
Designed a 5-stage pipeline for Relation extraction and classification data generation, along with a dataset spanning 255 relation types, with ~200k sentences
Selected in the top 4% nationally to attend the Amazon MLSS 2022
Selected in the top 1% nationally as for a 6 month long mentorship in quant research and finance by JPMorgan Chase
Selected in the top 32 of 6200 women nationally (top 0.5%) for the 6 moth long STEM initiative by D.E.Shaw
As an ML intern for 3 months at Caliche, I worked on the identification and classification of defects on the surfaces of steel sheets. I split the task into two: segmentation and classification!
Segmentation: I implemented a custom UNet Backboned architecture.
Classification: I leveraged Residual Neural Networks with skip connections and ReLU activation to achieve a 94% validation accuracy
The framework was integrated into a GUI for easy accessibility (e.g. outputs as shown in the last 2 images) !
See the implementation here!
In an interesting course on Compilers at IIT Guwahati, I implemented a target code translator from a 3-address code array with a symbol table, supporting auxiliary data structures to assembly language for NanoC.
The generated assembly code could then be translated with gcc assembler to produce executables, using machine-specific translation.
I used Flex for tokenisation and Bison for semantic analysis.
I secured an AS grade (awarded for outstanding performance to 1-2 people in the batch) for this bonus project!
I experimented with xv6 operating system and had the chance to implement some interesting X
Implemented a POSIX-like API to add kernel threads, and basic synchronisation primitives like spin-locks and mutexes.
I also implemented a hybrid scheduler using a mix of Shortest Job First scheduling and Round Robin scheduling.
Further to optimise xv6's memory management, I implemented lazy allocation, swap-based paging, thus reducing memory overhead and improving system performance!
Check out the implementation here.
Annotations take up a lot of monetary resources and man-hours. We were curious to predict the 'minimal' number of annotators required for a task concerning noisy data!
We wrote a short paper on our newly introduced Annotation Error Detection (AED) framework to determine the minimum number of annotators needed for reliable labelling in NLP datasets.
We applied our pipeline to toxicity classification of social media comments, detecting labelling inconsistencies and optimising annotation effort.
Majuli is the world's largest river island and the cultural capital of Assam, India.
We built an immersive 3D tour for the river island using Unity and Oculus to protect the rich cultural heritage of Majuli!
Check out the implementation here!
I have been fortunate to work with many Professors and research labs, along with industrial internships, that led me to this diverse skill set, which I seek to keep expanding!
As a part of the Computer Science and Engineering Association , the central student body at IITG's Computer Science department, I mentored a group of 10 freshers, helping them adjust to the cultural and academic environment. I guided them on internships, research projects and placements in addition to academic coursework.
As the Coordinator at Coding Club, I organised various ML & DL hackathons, led projects and mentored freshers on ML, AI, and Web/App Development.
I was one of the leads at the Core Team in IITG's AI club - IITG.ai. Here, I organised multiple research discussions and AI competitions. I also partook in organising guest lectures in NLP, CV and AI.
As the Lead of the AI team at IITG 4i Labs, I worked on various projects, including the speech setup and chatbot for a humanoid - Raman! I worked with people from diverse departments including mechanical and robotics students.
At JPMorgan Chase's Quant Mentorship Program (QMP), I led the Campus Marketing for IIT Guwahati. The program focuses on mentoring the best batch of sophomores at top schools in India on Quant. I myself was one of the selected mentees in my sophomore year; and finally joined the program back as a mentor when I joined JPMC full time!
I volunteered at Making A Difference, an NGO Foundation, to teach basic mathematics to over 300 tribal students at Palghar, a small district in rural Maharashtra. As an undergraduate at IIT Guwahati, I joined the National Service Scheme at to help as a teaching volunteer for underprivileged students during on-site rural visits.
+ 91 9819065254
voramahek21@gmail.com
Free AI Website Software