KAUSHIK BHAMIDIPATI

Software Engineer — Python / C++

SPEC_REF: KB-2026-PORTFOLIO
STATUS: ACTIVE_CANDIDATE|REV: 14.2.15
01 //

Professional Summary

Software engineer with an MEng in Computer Science and experience developing backend services, full-stack applications, retrieval systems, and data-intensive platforms. Strongest in Python and C++, with professional Java and Spring Boot experience and working knowledge of React and TypeScript. Experienced in building REST APIs, document-processing pipelines, cloud-cost analytics, automated tests, and interactive data products.

02 //

Technical Skills

Languages

PythonC++JavaSQLTypeScriptJavaScriptHTMLCSS

Backend

FastAPISpring BootREST APIsSQLAlchemyPydantic

Frontend

ReactStreamlit

AI & Data

Retrieval-Augmented Generation (RAG)LangChainChromaDBTensorFlowscikit-learnPower BI

Testing

pytestJUnitGoogleTest

Cloud & Tools

DockerGitGitHubAWSMicrosoft AzureGoogle Cloud Platform
03 //

Professional Experience

Software Engineer Intern

TidalWave
Remote — Hyderabad, India•2021–20226 months
  • Developed reusable REST API components and backend business logic using Java and Spring Boot within a microservices-based application.
  • Implemented authentication and authorization controls with Spring Security, separating protected application operations by authenticated user access.
  • Wrote and maintained JUnit tests, investigated backend defects, and validated application behavior before submitting changes for review.
  • Used Git-based development workflows to implement features, resolve defects, incorporate engineering feedback, and maintain changes across application components.
04 //

Selected Engineering Projects

Multi-Cloud Cost Optimization Platform

GitHub
PythonFastAPISQLAlchemyReactTypeScript
  • Built a full-stack cloud-cost analytics platform with a React and TypeScript interface and FastAPI/SQLAlchemy backend, normalizing provider-specific AWS, Azure, and Google Cloud records into a common data model.
  • Exposed four REST endpoints for service health, aggregate spending, provider comparisons, and workload-placement recommendations.
  • Developed a recommendation engine that considers alternative placement when estimated savings exceed 15% and CPU utilization differs by no more than 10 percentage points.
  • Implemented normalization logic for provider-specific cost, usage, region, resource, and performance fields, enabling equivalent workloads to be compared through a unified API.
  • Verified normalization and recommendation behavior through seven passing backend test cases, including a scenario identifying 40% estimated savings between comparable workloads.
  • Developed React views for spending summaries, provider comparisons, and optimization recommendations, separating data-fetching hooks from reusable presentation components.

RAG Document Intelligence

GitHub
PythonFastAPIStreamlitLangChainChromaDB
  • Built an offline-first retrieval application with seven API routes supporting PDF, Markdown, and plain-text ingestion, semantic search, document management, health monitoring, and source-level citations.
  • Implemented configurable document chunking, metadata preservation, local SentenceTransformer embeddings, and persistent ChromaDB indexing.
  • Added MD5-based duplicate detection to prevent repeated indexing and generated deterministic identifiers for stored document chunks.
  • Supported configurable top-k retrieval from 1–10 passages and two answer-generation modes: local extractive answering and optional LLM generation through OpenAI or Gemini.
  • Returned citations containing source filename, page number, chunk identifier, relevance score, and retrieved text, allowing users to verify answers against original documents.
  • Created seven automated tests covering file extraction, invalid and empty-file handling, chunk metadata, sentence splitting, health monitoring, duplicate prevention, retrieval, and database-reset behavior.
  • Developed a Streamlit interface for document upload, retrieval configuration, question answering, citation review, and application health visibility.

CacheCraft Concurrent Key-Value Store

2026–Present
C++20CMakeGoogleTestTCP Sockets
  • Developing a multithreaded in-memory key-value server in C++20 with five TCP commands: GET, SET, DELETE, EXPIRE, and STATS.
  • Implementing a bounded worker pool and synchronized storage layer to process concurrent client requests while protecting shared key-value and eviction state.
  • Adding per-key TTL expiration, configurable LRU eviction, snapshot persistence, and graceful server shutdown.
  • Building GoogleTest coverage for command parsing, concurrent reads and writes, expiration behavior, eviction ordering, snapshot recovery, malformed requests, and shutdown behavior.
  • Creating a reproducible benchmark matrix across 1, 8, and 32 concurrent clients to measure throughput, median latency, p95 latency, and error rate.
  • Structuring the server into independently testable networking, protocol, storage, eviction, expiration, persistence, and concurrency components.

Retail Customer Lifetime Value Analytics

GitHub
Pythonpandasscikit-learnStreamlitPower BI
  • Built a reproducible analytics and machine-learning pipeline processing 541,909 retail transactions through validation, cleaning, feature engineering, segmentation, model evaluation, and dashboard export.
  • Removed 135,080 records without customer identifiers, 5,268 duplicate transactions, and 2,517 records with invalid or non-positive pricing before generating customer-level features.
  • Converted the cleaned transaction history into 3,359 customer profiles containing recency, frequency, revenue, order-value, return-rate, product-diversity, and tenure features.
  • Prevented temporal leakage by separating a 274-day observation period from a 100-day prediction horizon before model training and evaluation.
  • Compared Random Forest and Gradient Boosting regressors against mean and historical-revenue baselines; the strongest baseline achieved an R2 of 0.691, MAE of GBP 467.92, and RMSE of GBP 1,040.27.
  • Identified a Champion segment representing 22.3% of customers but generating 67.8% of historical revenue.
  • Identified 201 at-risk customers associated with approximately GBP 247,500 in historical revenue and generated segment-specific retention recommendations.
  • Exported transaction, customer, product, and calendar datasets into a Power BI-ready star schema and created a Streamlit dashboard for segment, revenue, and retention analysis.
05 //

Education

University of Cincinnati

Cincinnati, Ohio•December 2025

Master of Engineering in Computer Science

SRM Institute of Science and Technology

India•May 2024

Bachelor of Technology in Computer Science and Engineering — Gaming Technology