Zishan Sami
zishansami102@gmail.com

Email  /  Resume /  GitHub  /  Twitter  /  Linkedin

I am Zishan, a 2020 graduate(B.Tech. + M.Tech.) from IIT Kharagpur and currently working full-time as a Smart Contract Engineer at prePO, a decentralised trading protocol for pre-IPO and pre-Token projects. Prior to joining prePO, I was working at Oracle Bangalore fulltime as an Applications Engineer and was parallely working part-time as a freelance web3 developer with intentions to gain experience.

I started my web3 journey last year with a scholarship from Coinbase for participating in DappCamp, an Ethereum Bootcamp by Preethi Kasireddy. I then moved on to do several freelancing project starting from building an ethereum based dapp end-to-end which does auto-payments for recurring subscription to working for SuperteamDAO where I helped several projects airdrop Solana NFTs to their community. For doing more serious work I participated in Secureum Bootcamp Epoch 0 (ethereum security and auditing bootcamp) and completed my CARE Audit for Sushiswap Bentobox. I also worked as a Technical Coach at DappCamp for cohort-2 in January. Right now I am fulltime at prePO as Smart contract Engineer.

Earlier during my college period, my experience and effort revolved around the areas of machine learning and deep learning and it's applications in computer vision & natural language processing. In the summer of 2019, I worked at IBM Research as an AI Research Intern in their Industrial AI Applications team working on deep learning methods for field boundary identification & using satellite images. Prior to this, I interned at Capillary Technologies in summer 2018 as a Machine Learning Intern where I worked on real-time Person Re-Identification problem and it was published in ICVGIP 2018.





Experience
prePO DAO
Smart Contract Engineer
Remote  • April 2022 - June 2022   • Proof of Work

• Implemented contract to claim staked token rewards using merkle tree proof. • Completed and deployed contracts for AcquisitionRoyale mini-games (an NFT based corporate battle game) to Polygon. • Implemented an upgrade over Ownable contract from OpenZeppelin called SafeOwnable to make it safe to transfer ownership. • Regularly reviewed contracts from other developers in the team. • Tech. Stack included Solidity, Hardhat, Node, Typescript, Mocha/Chai.

Oracle
Applications Engineer
Bangalore, India  • September 2020 - Now  

Working on building Rewards Service system for Oracle Cloud. Tech stack includes Java, Spring-Boot, Apache Kafka, Oracle DB.

Sumaiya Consultancy
Solidity Developer
Remote  • October - December 2021   • Demo Video   • Product Link (Ropsten Only)

Developed a decentralised protocol for recurring crypto payments such as monthly or yearly subscriptions. Deployed the protocol contract to Ethereum Ropsten Network and integrated it to a UI dashboard built in React. Tech. Stack included Solidity, Hardhat, NodeJs, Mocha/Chai and React.

DappCamp
Technical Coach
Remote  • January - February 2022  

Mentored and guided 4 teams from cohort-2 in their bootcamp projects who were new to the Ethereum development. Contributed in their project ideation, designing system architecture and resolving any technical roadblocks.

IBM Research
Summer Research Intern
Bangalore, India  • May - August 2019   • Poster Link

The work done during this period is now a part of a patent filed by IBM in US. Objective was to develop deep learning models for agricultural field boundary segmentation using multi-spectral and multi-temporal satellite images. I designed a deep Spatio-Temporal Segmentation Network architecture (ST-Net) which can extract both temporal and spatial features of a region from the time series of its SAR snapshots and RGB images leveraging the phenology timeline of agriculture crops. ST-Net uses a combination of Pixel-wise 3D LSTMs and DeepLabV3+ for temporal and spatial feature extraction, respectively. Achieved an improvement of 6% in mIOU compared to its existing counterparts.

Capillary Technologies
Machine Learning Intern
Bangalore, India  • May - July 2018   • Demo Link

Worked on Person Re-identification (ReID) problem to identify / re-identify people in real time and tag them with unique labels. Proposed a new loss function called Cluster Loss for deep CNN architectures motivated from triplet loss. Cluster Loss learns feature representations of people in a class independent way, forming round and separated clusters of all the classes. Improved the Rank-1 accuracy by 3% and mAP by 6% on two popular datasets.

Hackathons & Bootcamps
Secureum | Ethereum Security and Auditing Bootcamp
Ranked 50/1024 • October - December 2021   • Audit Report   • CARE NFT Badge

Completed 8 week long learning phase which covered Ethereum 101, internals of EVM, advanced Solidity, security best practices, audit tools and techniques and CTFs. Was selected (Ranked 50/1028) to perform audit on Sushiswap Bentobox as part if its CARE initiative. Got Ether NFT Badge for the audit report submitted.

DappCamp | Ethereum Developer Bootcamp by Preethi Kasireddy
Coinbase Scholarship Recipient • August 2021   • Product Repo

Built a decentralised app which allow brands to lock some rewards in a contract and automatically distribute it to people promoting the brand content on twitter using Chainlink

Microsoft Code.Fun.Do 2017
Top 3 Team • January 2017   • Demo Link

Secured runner up position in the winner list out of 20+ participating teams at IIT Kharagpur.
We developed a messenger bot called Chhuk-Chhuk which can respond to people on railway queries like the status of trains, their routes, PNR status, etc. Built a data scraper for real time railway information & integrated multiple APIs for language understanding capabilities. Full application backend was built on PHP.

Capillary Data Science Challenge 2017
Runner Up • September 2017   • GitHub Repo

We secured second position in the Machine Learning category out of 50+ participating teams.
We developed a movie recommendation system which can suggest movies to the users based on their ratings history on the mobile application. We did it by implementing a state of the art recommendation model called Biased Matrix Factorization on 1-M Movielens dataset using Tensorflow and integrated it on Flask backend with UI on android app.

OpenSoft 2019, Technology General Championship
Winner • March 2019   • Report Link

Designed a natural language search engine system for lawyers, legal experts and ordinary people where they can search for legal acts and previous court cases more effectively. Integrated submodules like Case Summarizer, Spelling Corrector, Query type detector, etc. Used InferSent as our sentence embeding model to rank sentences with respect to the input query. Developed an effective system design for reducing the search domain to small search space for fast ranking.

Safety Data Analytics Challenge 2017
Winner • September 2017   • GitHub Repo

Developed a machine learning model which can minimize the cost of customer retention & acquisition of a telecom company by accurately predicting its probable customer churning.

Open-Source
CNN-from-Scratch
GitHub     220+     70+
• Repository     • Live Demo

A simple CNN implemented in Numpy. It is currently the most popular repository for the ground up implementation of CNNs. Demo running on Flask server.

First-Impression
GitHub     20+     20+
• Repository    

A solution to the problem "First Impressions" given in ECCV’16. It is a TensorFlow implementation of a paper from ECCV’16 which can predict the big five personality traits of a user from their short video.

Research & Publications
Cluster Loss for Person Re-Identification
Doney Alex, Zishan Sami, Sumandeep Banerjee, Subrat Panda
Published in ICVGIP 2018   • arXiv Link    • Demo Link

Abstract : Person re-identification (ReID) is an important problem in computer vision, especially for video surveillance applications. The problem focuses on identifying people across different cameras or across different frames of the same camera. The main challenge lies in identifying the similarity of the same person against large appearance and structure variations, while differentiating between individuals. Recently, deep learning networks with triplet loss have become a common framework for person ReID. However, triplet loss focuses on obtaining correct orders on the training set. We demonstrate that it performs inferior in a clustering task. In this paper, we design a cluster loss, which can lead to the model output with a larger inter-class variation and a smaller intra-class variation compared to the triplet loss. As a result, our model has a better generalization ability and can achieve higher accuracy on the test set especially for a clustering task. We also introduce a batch hard training mechanism for improving the results and faster convergence of training.

Enabling Continual Learning by overcoming Catastrophic Forgetting
Bachelor Thesis | Guide : Pabitra Mitra
Computer Science Department  • Aug'18 - Apr'19   • Thesis Link

Abstract : Catastrophic Forgetting in deep neural networks is a term used in Continual Learning literature to signify the drastic performance drop on earlier learned tasks while learning a new task on the same model. The focus in continual learning is to minimise the catastrophic forgetting as much as possible without having access to the previous datasets. In this work we try to study the extent of catastrophic forgetting and how to overcome them while learning multiple tasks sequentially. We have implemented several popular methods and have presented a detailed comparitive analysis between all the methods on CIFAR100 dataset, by splitting it into 5 separate independent tasks. We have also tried and tested merging two methods together and concluded that our best performing combination is Lwf[8] and EWC[7] combined.

Leadership Roles
UC San Diego
Captain, OpenSoft Team 2018-19
Radhakrishnan Hall of Residence, IIT Kharagpur

Led the Gold Winning team of 15+ members from R.K. Hall of Residence in the Inter Hall OpenSoft Competition 2018-19.

Technology Advisor, 2018-19
Radhakrishnan Hall of Residence, IIT Kharagpur

Advised a pool of 200+ students from R.K. Hall of Residence for the prestigious annual Technology General Championship 2018-19.




© 2019 Zishan Sami. All rights reserved.
Website cloned from here.