VK Varshith Kaki
Resume
GitHub LinkedIn Email

Open to Software Engineering & AI/ML roles

Vishnu Sai Varshith Kaki

I build systems that scale, and AI that makes a difference. Backend platforms, distributed systems, and production LLM applications, shipped with the same rigor as the research behind them.

Currently MS Computer Science, University of Florida · 4.0 GPA
Background ~3 years production software engineering, Couture AI
Explore My Work Get in Touch

01 · Selected Work

Things I've shipped and broken
things to understand better.

ShellHacks / MLH 2026

Winner: [MLH] Best Use of Snowflake API

Hurricane Week

An interactive hurricane-preparedness simulator: you're a Miami-Dade resident with $400 and five days before landfall, and every decision (evacuate or shelter, insure or risk it) changes your financial and safety outcome. A deterministic rules engine drives every consequence; an AI coach (Gemini, grounded in your actual choices) explains the outcome afterward instead of deciding it.

  • Grounded Readiness Assistant answers real prep questions from verified FEMA/NOAA facts via Snowflake Cortex RAG, not generic LLM guesses
  • Replay mode diffs your new run against your last one and explains which decisions moved the score
  • An explicit AI-transparency panel discloses exactly what each AI feature receives, and no account or identity data ever leaves the system
  • Full stack: Next.js 16 + FastAPI, with MongoDB Atlas, TimescaleDB (Tiger Data), Snowflake, Gemini, ElevenLabs, and a Solana devnet completion record
Next.jsTypeScriptFastAPISnowflake CortexGeminiElevenLabsMongoDB Atlas
Devpost Source

Conversational Search Engine

Copart Conversational Search

Natural-language search and multi-turn refinement over ~26,000 real Copart auction lots. Ask for what you want in plain English, change your mind mid-conversation, and always see exactly what the system currently believes you asked for, backed by a deterministic, rule-based filter extractor with an optional LLM extractor behind the same interface.

  • Runs with zero API key: default extraction is deterministic and fully offline-testable
  • Surfaces a real data-coverage gap instead of hiding it: Copart's public export withholds price data, so the app detects that from the data itself and tells the user rather than estimating a number
  • Structured filtering with deterministic SQL retrieval and explicit conversational state across turns
PythonFastAPIReactSQLiteRule-based NLU
Source

Event-Driven Platform

RetailBrain

Event-driven retail intelligence platform streaming live sales events through Kafka/Redpanda, with a LangChain/LangGraph AI layer that answers business questions in natural language over the live data, and Prometheus/Grafana observability for latency and anomaly rates.

FastAPIKafkaPostgreSQLLangGraphDocker
View source

Real-Time Analytics

CryptoPulse AI

Real-time crypto analytics platform computing technical indicators over streaming market data. A grounded agent, sharing one tool registry with its own MCP server, explains market movements and flags anomalies, with JWT auth and a fully Dockerized local deployment.

ReactTypeScriptFastAPIKafkaMCP
View source

Systems Programming

NetPulse

C++17 TCP/UDP communication system: multithreaded connection handling, message routing, structured JSON logging, live throughput/latency metrics, and a UDP heartbeat discovery beacon, built for systems-and-networking interview prep.

C++17SocketsMultithreading
View source

3D Web

ModelVista 3D

Full-stack 3D product viewer for cataloging and inspecting GLB/GLTF assets from any URL-ready storage. Search and filter a catalog, then orbit/pan/zoom a model with React Three Fiber before it ships to ecommerce, AR, or configurator workflows.

React Three FiberFastAPIPostgreSQL
View source

Microservice

Resume Parser Service

A hosted resume-parsing microservice exposing a single parse endpoint, deployed to an Azure VM for an internal proof of concept during early internship work.

PythonREST APIAzure
View source
More from the archive: Gossip & Push-Sum, IIITS Grader, BookMySeva, and earlier builds
Gossip & Push-Sum Gleam/Erlang actor simulator for UF's Distributed Operating Systems Principles coursework. IIITS Grader Static site for calculating semester grades at IIIT Sri City. BookMySeva Django service-marketplace app, undergrad team capstone (5-person team). More on GitHub Full history of commits, experiments, and in-progress work.

02 · About

From IIIT Sri City to production
systems at Couture AI, now at UF.

I started out building full-stack class projects at IIIT Sri City, graduated into nearly three years as a software engineer at Couture AI, and I'm now back in school at the University of Florida to go deeper on the systems side: distributed systems, ML, and the theory behind the tools I'd already been shipping with in production.

Most of what I build sits at the intersection of backend engineering and applied AI: services that have to stay correct and fast under real load, with an LLM or retrieval layer bolted on in a way that's observable instead of magical. I like projects that are honest about their own limitations. Copart Conversational Search tells you outright when a filter can't match anything instead of faking a result, and that's the kind of engineering judgment I try to bring everywhere.

Outside of coursework and internships, I build weekend and hackathon projects (Hurricane Week, built for ShellHacks/MLH) because shipping something end-to-end in a weekend is still the fastest way I know to learn a new piece of infrastructure.

What I keep coming back to

  • Backend services & distributed systems
  • RAG, agents, and grounded LLM applications
  • Data platforms & streaming pipelines
  • Cloud infrastructure & observability
  • Hackathons & fast, scoped prototypes

03 · Experience

Nearly three years in production,
now split across research and industry.

  • Healthcare AI research: data cleaning and extraction, and NLP/ML model development for clinical decision support.
  • Built agentic AI workflows and automation for the NaviGator AI platform (n8n, LangChain, OpenAI APIs); created reusable prompt libraries and GenAI evaluation pipelines.
  • Fine-tuned AI responses on domain-specific datasets and ran adversarial and regression testing to improve relevance, accuracy, and reliability.
  • Architected end-to-end RAG systems with vector databases for context-aware healthcare document retrieval.
  • Built an AI-powered RCM Highlights Agent (LLMs + workflow orchestration) extracting insights from Revenue Cycle Management data and generating automated stakeholder-ready reports.
  • Benchmarked LLM frameworks (LangChain, LlamaIndex) on accuracy, latency, scalability, and cost.
  • Developed production LLM-powered microservices (summarization, classification, structured extraction) with prompt templates, output schema validation, and logging for reliable deployments.
  • Built a LangChain RAG system with FAISS vector search as a Dockerized inference API with Prometheus/Grafana monitoring; engineered LangGraph multi-agent workflows with guardrails and fallbacks.
  • Designed large-scale ETL pipelines (Azure Data Factory, Databricks) for pre-processing high-volume datasets; optimized runtimes via modularization and parallelism.
  • Revamped the Python backend of a global banking system (encryption, parallelization, modularization), cutting API latency 80%+; built Django/Flask REST APIs serving millions of daily requests.

04 · Skills

The stack I actually ship with.

AI / ML

LangChainLangGraphLlamaIndexRAG Prompt EngineeringLLMsPyTorchNLP

Languages

PythonGoJavaC / C++ TypeScriptSQL

Frameworks & Data

DjangoFlaskFastAPIReact KafkaSparkDatabricksPostgreSQL MongoDBRedis

Cloud & DevOps

AWSAzureGCPDocker KubernetesCI/CDPrometheusGrafana

05 · Education & Certifications

Credentials, briefly.

Aug 2025 – Dec 2026

University of Florida

M.S. Computer Science · GPA 4.0 / 4.0

Gainesville, FL

Aug 2019 – May 2023

IIIT Sri City

B.Tech Computer Science · CGPA 9.32 / 10

Sri City, India

Certifications

  • NVIDIA Certified: AI Agents (Multimodal)
  • DeepLearning.AI: Machine Learning Specialization
  • Microsoft Certified: Azure Fundamentals

06 · Contact

Have something interesting
in mind? Let's build it.

Hiring for software engineering, backend/distributed systems, or AI/ML roles, or just want to talk shop about an architecture? My inbox is open.

Resume
GitHub LinkedIn Email