Pushkar's
PORTFOLIO

I'm a rising senior passionate about integrating different scientific disciplines to discover something extraordinary.

See My Work

Curriculum Vitae

THINK DIFFerent

[ About me

I earned my bachelor's in technology at IIT Guwahati. I’ve conducted extensive research and taken rigorous courses spanning data science and AI/ML to gain a different perspective on my interests. I believe in creating an impact through my work and supporting and bettering daily lives. I'm also a skilled vocalist, guitarist, gamer, and artist.

See My Work

See My Work

Co-first author – Submitted to AI in Neuroscience journal

PIGMENT: A deep learning framework for Porcine Immunohistochemistry seGMENTation
Pushkar Ambastha, Javid Dadashkarimi, Sai Krishna C. Annavazala, Drew Parker, Ramon Diaz-Arrastia, Hailong Song, Rebecca P Donahue, Douglas H. Smith, Jean-Pierre Dollé, Victoria E. Johnson, John A. Wolf, Ragini Verma

PIGMENT: A deep learning framework for Porcine Immunohistochemistry seGMENTation
Pushkar Ambastha, Javid Dadashkarimi, Sai Krishna C. Annavazala, Drew Parker, Ramon Diaz-Arrastia, Hailong Song, Rebecca P Donahue, Douglas H. Smith, Jean-Pierre Dollé, Victoria E. Johnson, John A. Wolf, Ragini Verma

LINK

LINK

Submitted to ACM Computing Surveys

Pushkar Ambastha, From AlphaFold 2 to AlphaFold 3: A Review on Advancements in Protein Structure Prediction (Submitted to ACM Computing Surveys (Impact Factor: 23.8). Currently under review)

Investigated recent advances in protein structure prediction, like AlphaFold 3, which depicted a pattern toward the generalization ability of the models leading toward Large Language Models (LLMs).

Pushkar Ambastha, From AlphaFold 2 to AlphaFold 3: A Review on Advancements in Protein Structure Prediction (Submitted to ACM Computing Surveys (Impact Factor: 23.8). Currently under review)

Investigated recent advances in protein structure prediction, like AlphaFold 3, which depicted a pattern toward the generalization ability of the models leading toward Large Language Models (LLMs).

LINK

LINK

Featured Works

Research

FEATURED

Synthetic Human Memories

WHERE

MIT Media Lab

YEAR

2024-25

FEATURED

Synthetic Human Memories

WHERE

MIT Media Lab

YEAR

2024-25

FEATURED

Synthetic Human Memories

WHERE

MIT Media Lab

YEAR

2024-25

FEATURED

PIGMENT

WHERE

Univ. of Pennsylvania

YEAR

2024-26

FEATURED

PIGMENT

WHERE

Univ. of Pennsylvania

YEAR

2024-26

FEATURED

PIGMENT

WHERE

Univ. of Pennsylvania

YEAR

2024-26

FEATURED

Immunological ABM

WHERE

MIT Media Lab

YEAR

2023-25

FEATURED

Immunological ABM

WHERE

MIT Media Lab

YEAR

2023-25

FEATURED

Immunological ABM

WHERE

MIT Media Lab

YEAR

2023-25

See all the other works ↗

See all the other works

What I Work

Projects

Domain-specific Question Answering chatbot

We develop pipelines to retrieve a knowledge base article from the database based on the query and answer the query using the retrieved passage. We optimize the pipeline for performance, latency, and resource usage. Developed question-answering pipeline using techniques like model distillation, sparsification, pruning, and fine-tuning the DebertaV3-Base model to decrease inference time and have a minimum loss in accuracy. Project on PS given by DevRev.ai

We develop pipelines to retrieve a knowledge base article from the database based on the query and answer the query using the retrieved passage. We optimize the pipeline for performance, latency, and resource usage. Developed question-answering pipeline using techniques like model distillation, sparsification, pruning, and fine-tuning the DebertaV3-Base model to decrease inference time and have a minimum loss in accuracy.

We develop pipelines to retrieve a knowledge base article from the database based on the query and answer the query using the retrieved passage. We optimize the pipeline for performance, latency, and resource usage. Developed question-answering pipeline using techniques like model distillation, sparsification, pruning, and fine-tuning the DebertaV3-Base model to decrease inference time and have a minimum loss in accuracy. Project on problem statement given by DevRev.ai

GitHub
Report

Options Pricing and Market Calibration Simulator

Built a Black-Scholes options pricing simulator with Streamlit and yfinance, and modeled call or put Profit and Loss across 100+ spot-volatility grid scenarios. Generated 4 interactive heatmaps over a 10×10 volatility–spot grid to visualize PnL and pricing errors, identifying profitable scenarios and mispricings under shifting volatility regimes.

Built a Black-Scholes options pricing simulator with Streamlit and yfinance, and modeled call or put Profit and Loss across 100+ spot-volatility grid scenarios. Generated 4 interactive heatmaps over a 10×10 volatility–spot grid to visualize PnL and pricing errors under shifting volatility regimes.

Built a Black-Scholes options pricing simulator with Streamlit and yfinance, and modeled call or put Profit and Loss across 100+ spot-volatility grid scenarios. Generated 4 interactive heatmaps over a 10×10 volatility–spot grid to visualize PnL and pricing errors, identifying profitable scenarios and mispricings under shifting volatility regimes.

GitHub
Options Pricing App

BioNLP: Genomic and Protein Analysis with Transformers

Designed transformer-based pipelines, leveraging DNABERT for functional genomics with a 5% F1-score improvement using minimal fine-tuning on small datasets, and built a SARS-CoV-2 genome classifier achieving 99% accuracy, surpassing random forest by 40% in precision. Developed a Flask-based protein design and active site analysis platform, generating over 10,000 custom proteins, boosting ligand-binding prediction by 20%, and automating .pdb workflows.

Designed transformer-based pipelines, leveraging DNABERT for functional genomics with a 5% F1-score improvement using minimal fine-tuning on small datasets, and built a SARS-CoV-2 genome classifier achieving 99% accuracy, surpassing random forest by 40% in precision. Developed a Flask-based protein design and active site analysis platform, generating over 10,000 custom proteins.

Designed transformer-based pipelines, leveraging DNABERT for functional genomics with a 5% F1-score improvement using minimal fine-tuning on small datasets, and built a SARS-CoV-2 genome classifier achieving 99% accuracy, surpassing random forest by 40% in precision. Developed a Flask-based protein design and active site analysis platform, generating over 10,000 custom proteins, boosting ligand-binding prediction by 20%, and automating .pdb workflows.

GitHub - 1
GitHub - 2

Cover Generation using
OpenAI tools

Cover Generation using OpenAI tools

Developed a multi-modal pipeline that converts audio/text input into images using state-of-the-art OpenAI tools. Generated optimal transcripts for the podcasts and songs with OpenAI Whisper to use in creating prompts. Designed pipeline with Latent Diffusion Models (DALL-E) to generate aesthetic cover images from created prompts using ChatGPT/GPT-2 models. Project by IITG.ai Club, IITG

Developed a multi-modal pipeline that converts audio/text input into images using state-of-the-art OpenAI tools. Generated optimal transcripts for the podcasts and songs with OpenAI Whisper to use in creating prompts.

GitHub
HF-Spaces

Super Resolution Photo-Mosaic

Developed a Computer Vision pipeline that enhances the images by super-resolution and image stitching. Designed a multi-model pipeline consisting mainly of the Latent Diffusion Upscaler model for super-resolution and Image Stitcher to create a panorama.Github. Project by Coding Club, IITG


GitHub

Path

Four institutions, one throughline: rigorous problem-solving at the intersection of ML and research.

01

IITG

01

IITG

Undergraduate training in Engineering foundation and quantitative fields

02

MIT

02

MIT

Two labs, two years, built differentiable ABMs and ran statistical studies on AI-driven false memory formation.

03

UPenn

03

UPenn

Co-first author on a segmentation framework outperforming SOTA baselines by up to 95% detection rate.

04

Imperial

04

Imperial

Starting MSc AI Applications on a competitive scholarship — solving quantitative and machine learning problems

01

Undergraduate training in Engineering foundation and quantitative fields

IITG

02

Two labs, two years, built differentiable ABMs and ran statistical studies on AI-driven false memory formation.

MIT

03

Co-first author on a segmentation framework outperforming SOTA baselines by up to 95% detection rate.

UPenn

04

Starting MSc AI Applications on a competitive scholarship — solving quantitative and machine learning problems

Imperial

01

IITG

01

IITG

Undergraduate training in Engineering foundation and quantitative fields

02

MIT

02

MIT

Two labs, two years, built differentiable ABMs and ran statistical studies on AI-driven false memory formation.

03

UPenn

03

UPenn

Co-first author on a segmentation framework outperforming SOTA baselines by up to 95% detection rate.

04

Imperial

04

Imperial

Starting MSc AI Applications on a competitive scholarship — solving quantitative and machine learning problems

AWARDS

2026 Imperial India Future Leaders Scholar

Kaggle 4X EXPERT - 2023

2023 Inter IIT Tech Meet Gold Medallist

Bronze Medal (85th/1100), Kaggle - 2023

Bronze Medal (85th/1100), Kaggle, 2023

Bronze Medal (99th/1025), Kaggle - 2023

Bronze Medal (99th/1025), Kaggle, 2023

Convolve Hackathon (28th/231) - 2022

EXPLORE
MY SOCIAls

FACEBOOK

  • FACEBOOK

FACEBOOK

BEHANCE

  • BEHANCE

BEHANCE

INSTAGRAM

  • INSTAGRAM

INSTAGRAM

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

  • SAY HELLO!

Let’s work together!

[ Marked fields are required to fill ]

Let’s work together!

[ Marked fields are required to fill ]

Let’s work together!

[ Marked fields are required to fill ]

Create a free website with Framer, the website builder loved by startups, designers and agencies.