Avikalp Gupta

ODF is a selective founders' community (with a 5% selection rate).
As a part of the ODF23 cohort, I met ~100 exceptionally talented, ambitious and brave innovators from across the world. Every person in the cohort had either founded or supported the founding of at least one high impact product.

I had two major responsibilities as the head of the data science team of the company:
1. Team building
2. Development and delivery of ML features in our application

TEAM BUILDING
* Hiring: Over a period of 8 months, after evaluating 100+ candidates (each with more than a year of experience as a data scientist, after completing studies from one of the IITs), I was able to hire 2 absolute gems as the first data scientists of the company.
* Culture setting: From the very beginning, my team embodied the characteristic of infinite responsibility. Every person in the team was given a problem that the company is facing and empowered to build a solution to it using all the resources that the company could access.
* Collaboration framework: Our data science team followed the principles of software engineering for collaboration and version controlling in both, the research and development phases of the projects. We used tools like DVC and JupyText to set up the framework with enables us to contribute effectively, get peer feedback and reproduce results.

DEVELOPMENT & DELIVERY
* Fraud detection modelling:
- Extracted actionable insights from user behaviour patterns in wallet transactions, and their interactions with the platform.
- End-to-end pipeline to identify real-time fraud in Poker, clique detection using graph techniques.
- Saved around ₹8-12 lakhs per month of fraudulent money rotations and laundering.
* Poker bot:
- Created a bot mode in an open-source casual Poker project
- Led the creation of an AI bot that outperforms medium-level Poker players using Markov Chain modelling.
- Productized the AI bot into a learning assistant with features like adaptive difficulty and explanations of exploitable strategies.
* Friend request recommendation system:
- Led the creation of a recommender system using XGBoost that is 16% more efficient than the existing version getting friend requests accepted.

Co-founded InforMED along with Keaton Harris from South Afrika, Marcia Cunha Dos Santos from Brasil and Ishita Aggarwal from Canada for solving the communication problems faced by healthcare practitioners and their patients.

Job responsibilities:
1. App development (created a prototype of the application using Android Studio and an MVP using Ionic Framework)
2. Partnerships and Collaborations (Brought Medallion Social Innovators on board for POC and testing)
3. Executive responsibilities (Company structure, company registration, logistics etc.)
4. Marketing (Focused marketing to relevant individuals in events like Startup Master Class Pune, Social Entrepreneurship and Enterprises (SEE) Central in IIT Kanpur, Google Developers' Group (Pune) DevFest'18 etc.)

Selected as one of the top 21 candidates, out of around 9000 applicants, for the annual YSI Global program for creating sustainability solutions for achieving United Nation's Sustainable Development Goals (SDGs) 2030.

Worked on really cool Artificial Intelligence solutions for various domains, including but not limited to natural language processing, supply-chain optimization, computer vision, time-series predictions etc. Gained exposure to various industry verticals like entertainment, pharmaceuticals, fashion, clinical trials, manufacturing, etc.

Technologies and algorithms used:
1. Python Flask
2. Elastic Search
3. NLTK Python Library
4. Jupyter Notebook
5. Scikit-learn
6. Probabilistic Machine Learning (Markov Chain Model)
7. Pandas (+Numpy, Matplotlib, Seaborn etc.)
8. Python OpenCV Library
9. Keras
10. TensorFlow
11. Neural Networks (CNN, LSTM, I-GAN, etc.)
12. ARIMA, Holt-winters
13. Reinforcement Learning
14. Alexa Skills

* Migrated their entire codebase from Subversion to Git and hosted it on GitLab.
* Integrated e-Doctor services with the e-Mitra platform of the Government of Rajasthan
* Reviewed the company's compliance with the Electronic Health Records (EHR) Standards of India (2016).
* Implemented critical utility features in their service like Skype integration with the website and screen recording of the consultations for quality control

Technology stack: PHP, CodeIgnitor, MySQL, SVN, Ubuntu

Worked in the research lab in Bangalore (India) for a year.
• Leveraging ICT for Solid Waste Management (SWM) Mar’17-Oct’17
· Explored the SWM infrastructure & implementations for possible technological intervention
· Developed an Android app to track different categories of waste using GSM-based tracking devices in an effort to verify the claims of BBMP and to identify major dumping sites
· Lead the soft-tech solutions’ team at a startup on waste segregation (trashcon.in) for 6 months in 2017
• Joint Query+Resource Optimization Sep’16-Mar’17 · Studied the PerfOrator project and designed a method to combine the optimization steps
· Analyzed hardware modelling for clusters on Azure with HDFS, WASB & ADLS storage options
• Program synthesis Jul’16-Sep’16 Fabricated & implemented an optimized algorithm to find “diamonds” in a query-plan, to replace them by UDFs

As a part of a small team, we experimented with the latest research done by the scientific community on Visual Search (shallow networks, Siamese triplet networks etc.) to suggest the most relevant method & provide an optimized implementation for providing image-based search for e-commerce fashion websites.

Minimized Risk-weighted Assets (RWAs) for the company by making the allocation optimizer aware about it
• Enhanced the internal engine which recommends collateral allocation to secured funding trades to account for
trade-offs between liquidity considerations (maximizing liquidity) and capital considerations (RWAs)
• Quantified cost savings for the firm if the collateral was allocated to the trades based on the enhanced engine.
Skills Earned: Scalable programming in functional languages, Introductory knowledge in Finance

As it was a start-up, there were only 7 members in the team. My main responsibility was to create an adaptive learning practice session for the website and manage its databases and back-end code.
• Project Chameleon: Thoroughly studied the development, pros and cons of the mastery-based adaptive learning model of Khan Academy and implemented an algorithm apt for the preparation for competitive examinations (Skills Earned:- CakePHP, PHP, MySQL, Git, OOPS)
• Gamification (Integration with Backend): Integrated the adaptive backend model with the Gamified user interface using AJAX and self-created APIs in CakePHP and JS (Skills Earned:- HTML+CSS, JQuery, AJAX
• Database Designing: Made architectural changes in database to extend the functionality of platform so that
each user can prepare for multiple competitive exams (Skills earned: Database Architecture Proficiency)
• Sales and Product Pitching: Convinced 97 of 120 students to join the SaaS based platform