Divyansh Prateek

Project - Services Platform (SPLAT)

Module 1 - Resourcing and Recommendations

• SPLAT, internal resourcing tool, optimizes resources according to requirements and generates recommendations based on metrics like skill set, bench aging, and time to deploy.
• Designed and developed prototype showcasing usability. Gathered and accumulated organization-wide data. Orchestrated application showcasing stakeholders, product owners, and founders for feasibility.
• Gathering technical requirements from Business operations team to migrate from traditional sheet based operations to automated UI based resource allocations.
• Integration with workday to fetch resource, project and allocation data.
• Responsible for solutioning application infrastructure into multiple microservices for notification, authentication, allocation, job and marketplace. Built role based authorization for applications.
• Developed data pipelines to be utilized for analytical purposes. Developed key metrics based on OKR's defined by the product team. Achieved > 95% allocations in the platform.
• Implemented and optimized org wide resource metrics algorithms such as bench aging, time to deploy, priority booking, flag logic and overlap calculation.
• Packaged common DB schema npm package and common utilities for all microservices.
• Developed decentralized flow for allocation with skill feedback where tech leads and PMs are responsible for evaluating resource. Achieved OKR where 100% resources were allocated in decentralized flow.
• Optimized application response time by building pipelines for near real time transactions.
• Tech stack - Python, Cloud Functions, BigQuery, Node js, Cloud Build, Artifact Registry, Dataflow, Cloud Run, Pub/Sub, React js, Postgresql

Module 2 - Talent Exchange

• Utilizing resourcing module Talent exchange allocates resources based on availability and skill set to accomplish tasks for short duration (spike, POC, integrations).
• Led team of software developers, platform engineers and machine learning engineers to execute resource tasks across all business units.
• Built efficient scalable email notification system and chat bots for effective communication.
• Tech stack - Google chat API, Dialogflow, Node js

Project - Codexcelerate

Codexcelerate is an org wide initiative to build Generative AI based accelerators to enhance productivity of the delivery of projects.

Application 1 - Qodequre

• Developed debugging assistant leveraging generative AI that aid developers in resolving errors and exceptions.
• Deployed multiple packages for various programming languages and frameworks to aid development process.
• Built vs code extension to aid bug resolution through vs code.
• Tech stack - Fast API, node js, python, Vertex AI

Application 2 - QlosureEase

• Led team of software developers, machine learning engineers and other resources from marketplace to orchestrate application to generate automated unit test cases, docstring, code explanation and code overview.
• Researched on multiple multi agent frameworks such as meta gpt and chat dev for building such use case. Also explored on challenger and champions LLM models for code generation capabilities.
• Aligned team on validating code on multiple repos and achieved relevance percentage of more than 80%. Planned release cycle for application. Achieved excellent adoption of over 80%.
• Facilitated development of analytical dashboards using looker studio.
• Tech stack - Flask, Angular, Docker

Application 3 - Codex (SPLAT)

• Codex provides resource forecasting, SOW retrievals, acceleration and artifacts recommendations based on project description, tentative start month of project and resources core role as input.
• Engineered SOW retrieval by generating embeddings. Implemented RAG for SOW retrieval, LLM for text summarization and top k matches based on project description. Integrated other internal applications utilizing accelerators and artifacts.
• Facilitated development of analytical dashboards using
• Tech stack - Fast API, Langchain, python, Cloud Run, GCS

Project - Terra.do

• Educative platform for climate change and sustainable development.
• Developed analytics dashboard for events archive and course content module which includes students behaviour and segmented analytics using k-means clustering.
• Tech stack - Python, node js, data visualization

Project - Frejun/FrejunHR

• Frejun is AI based web application which automates recruitment process by screening resumes, shortlisting candidates, automating interview calls, recruitment and shortlisting, and better decision making.
• Key responsibilities includes building backend infrastructure using Django which included payments module, account management, meetings, scheduling and third party integrations. Developed microservices for audio conversion and processing, recording meetings by running automation script on containers which was deployed on AWS ECS. Developed text based classification using NLP to extract keywords related to job requirement.
• Tech stack - Django rest framework, GCP, AWS, docker, te

Project - Deal Street Asia [3 months]

• Web portal for financial news/feeds across Asia which includes venture capital and private equity news.
• Designed and developed corporate subscriptions which included licenses which were bought by organization. Licenses had expiry and could be reassigned to any user registered belonging to organization. Predictive analytics using SVM, Random forest, Logistic regression and XGBoost to predict churn for corporate licenses.
• Tech stack - Next js, Django rest framework

Project - Apnahood

• Apnahood is social shopping platform for communities where neighbours can recommend, buy and sell in a simple, fun and private environment.
• Developed MVP for community members which included hybrid mobile application for home entrepreneurs to schedule event and connecting local vendor connect to communities and sell their products and services. Created admin panel and automated tasks for operations team for better event and catalogue management. Setup analytics to track customer behavior. Collaborated with UX team for better user experience and simplicity. Used segmented analytics techniques to analyze customer responses and time spent in buying product.
• Tech stack - Flutter, Firebase, node js, AWS, retool

Project - Faster.Fit

• Faster.Fit is designed and developed to cater needs of professional athletes who are looking for trainers or similar communities.
• Developed UI for user profile modules, sending requests to trainers for scheduling sessions with athletes. Payments options for sessions for trainers and bank verification. Predictions for performance based analysis and suggestions using training score.
• Tech stack - Flutter, Firebase

Project – FAFM, IFRS, FINTL.

• Description - Transformation Layer (FAFM, IFRS, FINTL) is responsible for enrichments on inventories, accounting, payroll files by generating Accounting Entries and storing in data lake.
• Key Responsibilities - Responsible for developing RESTful API in Java and hibernate to interact with SQL.
• Developing modern web applications with excellent user experience.

Project – UNITETL, FINTL.

• Description - UNITE is web application which is responsible for performing enrichments for credits, loans and customer events.
• FINTL is responsible for enriching payroll information and inventories which is sent to rules engine (FARCORE).
• Key Responsibilities - Responsible for developing micro services using spring framework.

Project – FARCORE.

• Description - FARCORE is rules engine which enforces rules on Accounting Entries and Inventories.
• Files are sent to General Ledger for generation of invoices.
• Key Responsibilities - Provided production support using tools such as SNOW and Elastic Search.