Command Palette

Search for a command to run...

HomeAboutTrueStarProjectsExperienceSkillsEducationContactRésumé
~/portfolio $ whoami

Vineet Kumar

<ai_software_engineer />

AI software engineer building agentic, AI-native platforms on top of resilient distributed systems.

View Work
Get in touch
// 01 / about

About me

Last summer I interned at Microsoft on the platform behind Sentinel's data lake, building a workload-placement engine for about 7 million pipeline jobs. The part I'm proudest of wasn't code: I questioned an assumption the team had designed around, checked it against 30 days of production data, and found it was wrong.

Before that, three and a half years at ION Trading writing trade systems in C#, Java, and C++. Real users, real money, real 2am pages. It taught me that boring and well-tested beats clever.

Now I'm finishing a Master's at San Jose State, graduating May 2027, and co-founding TrueStar on the side. I'd rather work on something hard with people who know more than me than be the most confident person in the room.

  • LLMs / RAG
  • AI Agents
  • Azure OpenAI
  • TypeScript
  • React
  • Next.js
  • Three.js / GLSL
  • Node.js
  • Java
  • Python
  • C++ / C#
  • AWS
  • Docker
  • Kubernetes
Vineet Kumar
4+
Years in tech
14+
Projects shipped
120+
Modules automated
3.87
Undergrad GPA
// 02 / truestar

TrueStar

Co-Founder & Engineer

An AI-native expert research platform.

TrueStar runs two research tracks in parallel and merges them into one cited report. Primary research is live, AI-moderated expert interviews: a voice agent named Aria interviews vetted human experts over LiveKit, paying them per interview via Stripe Connect. Secondary research is TARS: four specialized AI agents that independently research the open web, debate their findings, cross-check claims, and produce a synthesis with a hallucination audit and grounding score.

Aria, the real-time voice interviewer

Sub-second voice agent over LiveKit WebRTC (Deepgram STT · Azure OpenAI · Fish Audio TTS) conducting 10-minute expert interviews.

TARS, multi-agent fact verification

Four adversarial agents (Data Scientist, Investigative Journalist, Domain Expert, Devil's Advocate) debate, reflect, and synthesize an auditable verdict with cited sources.

RAG document intelligence

Namespace-scoped retrieval over expert documents, embedded into Pinecone with Cohere Rerank 3.5.

Production microservices

Four services on a shared Docker network: React 19 SPA + Express API, FastAPI RAG, FastAPI multi-agent orchestrator, and the voice worker.

  • React 19
  • Express
  • PostgreSQL 17
  • Redis
  • Azure OpenAI
  • Pinecone
  • Cohere Rerank
  • Tavily
  • LiveKit
  • Deepgram
  • Stripe Connect
  • Docker
Visit truestar.techsource: private
// 03 / projects

Projects

Mostly grad-school and side projects, AI and distributed systems first. Filter by focus, or open a card to see what it actually does.

See more on GitHub
// 04 / experience

Where I've worked

Software Engineer Intern

·Microsoft
Redmond, Washington · On-site
May 2026 to Aug 2026
  • Designed and delivered an end-to-end intelligent workload-placement engine in C# and .NET that routes about 7 million shared-tenant pipeline jobs onto right-sized Kubernetes compute, engineered to raise fleet CPU utilization about 5x (11% to 59%) and cut compute cost.
  • Architected a two-step, config-driven placement algorithm with a network-boundary (VNet) match guard that structurally prevents mis-routing across isolated networks, shipped behind feature flags for zero-downtime, reversible rollout.
  • Led a data-driven design correction by analyzing 603K production runs over 30 days in Kusto (ADX) on Azure Data Lake, then hardened it with 22 integration tests and a classifier suite.
  • C#
  • .NET
  • Kubernetes
  • Kusto (ADX)
  • Azure Data Lake
  • Distributed Systems
  • Feature Flags

Software Development Engineer

·ION Trading
Pune, India
Jan 2022 to Aug 2025
  • Built a C# desktop interface to streamline import/export of market index, correlation, and volatility data in XML, cutting setup time from 1 week to 10 minutes between UAT and Production.
  • Engineered a trade-processing framework with Apache Camel and Java, integrating Kafka and ActiveMQ for ExxonMobil to enable real-time trade data exchange and reduce latency.
  • Enhanced a commodity trading module using Java Native Interface and C++ to automate bulk trade scheduling, cutting UAT time from 4 weeks to 30 minutes.
  • Automated .NET Framework upgrades with PowerShell across 120+ modules, slashing upgrade time from 2 months to 1 week.
  • Achieved a consistent 95% CI pass rate through rigorous code reviews, SonarQube, and SOLID-principles sessions for the team and new hires.
  • Authored technical documentation that reduced new-hire onboarding time by 60%.
  • C#
  • Java
  • C++
  • Apache Camel
  • Kafka
  • ActiveMQ
  • PowerShell
  • SonarQube
// 05 / skills

What I work with

Languages
  • Java
  • C#
  • C++
  • Python
  • TypeScript
  • SQL
  • Bash
AI & Retrieval
  • LLMs
  • RAG
  • AI agents
  • Prompt engineering
  • Embeddings
  • Vector search
  • Azure OpenAI
  • Pinecone
  • FAISS
Backend & Distributed
  • Microservices
  • gRPC
  • REST APIs
  • Spring Boot
  • FastAPI
  • Node.js
  • Apache Kafka
  • ActiveMQ
  • Apache Camel
  • Fault tolerance
Frontend
  • React
  • Next.js
  • Tailwind CSS
  • GSAP
  • Three.js
Data & Storage
  • PostgreSQL
  • Redis
  • MongoDB
  • MySQL
  • Prisma
  • Drizzle
Cloud & Delivery
  • AWS
  • Azure
  • Docker
  • Kubernetes
  • CI/CD
  • Linux
  • Nginx
  • OpenTelemetry
Practices
  • OOP
  • SOLID
  • Design patterns
  • Test-driven development
  • Code review
  • Agile
// 06 / education

Education, achievements, volunteering and mentorship

Education

  • Aug 2025 to May 2027
    Master of Science
    Computer Software Engineering
    San Jose State University · San Jose, California
  • Jul 2018 to Jun 2022
    Bachelor of Technology
    Computer Science & Engineering (Information Security)
    Vellore Institute of Technology · Vellore, India
    GPA 3.87

Volunteering & mentorship

  • Jan 2020 to Aug 2021
    Microsoft Learn Student Ambassador
    Microsoft · Beta rank
    Ran hands-on workshops for 150 students, every attendee shipping a full-stack deployment on Node.js, Docker, and Nginx, plus Azure sessions for 200+ participants.
  • 2019 to 2022
    Author, Code To Express
    Vellore Institute of Technology
    Wrote and maintained an open data-structures and algorithms repository used by 250+ students.

Certifications

  • Google Cloud Fundamentals: Core Infrastructure
  • Fundamentals of Parallelism on Intel Architecture
  • Cybersecurity and the Internet of Things
  • Microsoft Learn Student Ambassador

Achievements

  • Awarded Beta-rank Microsoft Learn Student Ambassador by Microsoft.
  • 2nd Place, Code Run Seek, IEEE IAS and IEEE WIE International Techno Carnival, VIT.
// 07 / contact

Let's build something

I'm graduating in May 2027 and open to new-grad software engineering roles, especially in AI and distributed systems. Always up for a good conversation or a side project.

vineetkia@gmail.com