You are on page 1of 9

PS 2 Mid Sem Report

Name: Snehil S Kumar


ID: 2019B2A30985P
Company: Propellyr Chaintech Inc

Introduction

I am currently interning at Propellyr Chaintech, an emerging fintech


company building innovative blockchain and AI-based financial
applications. This was an invaluable learning experience that gave me
exposure to cutting-edge technologies and real-world software
development.

In this extensive report, I will provide a deep dive into my three month
internship at Propellyr. First, I will describe Propellyr, its products and my
team. Next, I will explain my key internship project in detail including the
problem statement, technical architecture, implementation, and learning
outcomes. Further, I will summarise the diverse skills I developed across
software engineering, machine learning, blockchain, product design and
business domains. Lastly, I will reflect on the overall impact of this
internship experience on my career goals in the technology industry.

About Propellyr

Propellyr Chaintech is a fintech firm based in Delhi, India that was


started in 2022 by a team of few employees. Propellyr aims to leverage
emerging technologies to transform personal financial management and
investment advisory both in the space of stocks and cryptocurrencies. It
has recently gone under a merger with Yield 3 which is also developing
applications in the crypto data analytics space.
Propellyr pursues to develop digital platforms, mobile apps and
intelligent bots targeted for financial institutions to help them better
manage finances, investments, taxes and more. Their products combine
cutting-edge technologies like artificial intelligence, blockchain, and
analytics to provide personalised and automated finance solutions. Key
products and solutions developed by Propellyr include:

● Impact Calculator – A tool for market impact insights, portfolio


optimization, and strategy simulation, leveraging advanced
algorithms.
● Earnings Dashboard – A Real-time data platform for maximising
earnings, facilitating informed decisions through data-driven
strategies.
● Risk Management Dashboard – A sophisticated Web3 dashboard
for smart investments, tailored insights, and elevated risk
assessment capabilities.
● Yield Assistant – A personalised finance bot that acts as a virtual
financial advisor and money manager through natural language
conversations.

Internship Project – Building an AI Financial


Assistant

During my internship, I took up the exciting project of developing Yield


Assistant - an AI-powered financial assistant chatbot targeted at digitally
native users. Yield Assistant allows users to query financial data through
natural language conversations. My role involved conceptualising
features, architecting the system, implementing machine learning
models and integrating the backend platform.
Project Objectives

The goal of the project was to develop a working prototype of the Yield
Assistant financial assistant with capabilities like:

● Conversational interface for text-based interactions


● Personalised finance management recommendations using AI
● Financial Modelling of the portfolio assets
● Providing optimised allocations for the portfolio based on financial
models.

Technical Architecture

The system comprises two key components - the conversational


interface and the backend platform.

The conversational interface is built using Langchain, an open source


conversational AI framework. Langchain provides the natural language
processing pipelines for intent recognition, entity extraction and
embedding storages. The interface supports both text based
interactions.

The backend platform runs on AWS cloud and implements the core
functionality. It has an Apache Iceberg for storage of the data format and
Trino is used on the user side for querying the engine, while the other
architecture is built on python to run and scale the machine learning
models. Key components of the backend include:

● User and data management using SpiceDB


● Financial modelling and analytics
● ML models for recommendation and personalization
● Security using SpiceDB
● REST APIs to connect the interface with backend services
Additionally, the system uses AWS services like S3 , ECS, and Athena to
Comprehend for storage, data pipelines, modelling and analytics.

Data Flow

The conversational interactions are processed by Langchain to extract


semantic information like intents, entities from the user query and get
the context from the embeddings made using the Catalog files of
Iceberg. This data is passed to the backend platform through API calls.
The data is cleaned and processed in Spark and passed through ML
pipelines.

Models like recommendation, prediction and analytics are served in


real-time. The resulting insights are returned to the conversational
interface to respond to the user.

A similar architecture is shown below.

Key Technical Implementation

During my internship, I worked on the full technology stack to bring the


system together. Some key tasks included:
Frontend Conversational Interface

● Developed conversation flows for key user interactions like


expense adding, bill pay, budgeting etc.
● Implemented Rasa NLU pipelines for intent classification and
entity extraction. Built custom finance entities and intents.
● Created dialog models for natural conversation routing and context
management
● Integrated speech recognition and text-to-speech services for
voice capabilities

Backend Platform

● Built REST APIs in Node.js for user management, transaction


processing, account integration and model serving.
● Developed Ethereum blockchain integrations using web3.py for
secure financial data sharing.
● Implemented ML models in Python and TensorFlow for
expenditure prediction, budget recommendations and anomaly
detection.
● Created scripts for scheduled model retraining and deployment
automation using Jenkins and Kubernetes
● Setup Spark pipelines on AWS EMR for data ingestion,
preprocessing and feature engineering
● Designed MongoDB database schema optimised for financial
transaction data and analytics.

In addition, I defined product requirements, created UI/UX flows, and


contributed to design discussions to improve the end user experience.

Key Learnings

Through this project, I gained extensive hands-on experience in


cutting-edge technologies as summarised below:
Conversational AI: Worked extensively with Langchain to build
production-grade virtual assistant capabilities like accurate intent
classification and natural language generation.

Machine Learning: Gained hands-on ML experience for NLP


recommendation systems, anomaly detection and model optimization.

Backend System: Build on Python having modules for risk calculation


and portfolio management APIs along with a separate AI module for
catering interactions with the user

In addition to the technical skills, this project enhanced my


understanding of fintech product development, UX design and the agile
software methodologies used at Propellyr.

Deployed the project live on Propellyr Data Platform


Broader Internship Experience

The Yield Assistant project was the primary focus of my internship.


However, I also got opportunities to gain broader experience in software
engineering, machine learning and blockchain technology.
Collaborating with other teams on projects like fraud prediction,
customer segmentation, modelling research and API building expanded
my exposure to different technologies and approaches used at Propellyr.
Some key areas where I developed deeper skills through the internship
are:

Software Development Lifecycle: I gained end-to-end exposure in agile -


from requirement analysis, design and development, deployment and
monitoring.

CI/CD and DevOps: Worked with tools like Git, Kubernetes and Docker for
version control, continuous integration, infrastructure automation and
delivery pipelines.

Product Design: Got opportunities to work on user research, experience


mapping and prototyping during feature brainstorming which enhanced
my product design skills.

Quantitative Finance: Improved my financial modelling and analytical


skills through feature development for investment insights, cash flow
analysis and financial planning use cases.
https://www.propellyr.com/company

Key Takeaways from Internship

The internship at Propellyr provided me with an invaluable learning


experience and helped me grow both professionally and personally. Here
are some of the key takeaways for me from this fruitful internship:

● Developed hands-on specialised skills in cutting-edge technologies


like conversational AI, blockchain, and cloud-based Technologies.
● Got end-to-end exposure in building real world software products
using agile methodologies.
● Improved core technical skills in full-stack development, data
science and infrastructure automation.
● Built expertise in fintech product development including UX design,
financial modelling and domain knowledge.
● Experienced the culture and workings of a fast-paced technology
startup.
● Developed business acumen through cross-functional
collaboration and product innovation processes.
● Strengthened soft skills like communication, time management,
prioritisation and teamwork.
● Expanded professional network and industry connections.

Conclusion
In conclusion, my internship at Propellyr Chaintech provided immense
learning and growth. The hands-on experience in building an AI financial
assistant using bleeding edge technologies was truly rewarding. I am
grateful to have received guidance from talented engineers and business
leaders at Propellyr.

The knowledge and skills gained during this internship have expanded
my capabilities and prepared me well for graduate studies and career
opportunities at leading technology and fintech firms. I am excited to
apply my expanded expertise in future roles and make meaningful
contributions to product development and innovation. This internship
has given me confidence in my abilities to take on challenging real-world
projects and thrive in agile team environments.

You might also like