# Vikranth Reddimasu

> AI Engineer building agentic AI systems end-to-end. Open to work.

## About

Vikranth Reddimasu is an M.S. Data Science student at the University of Maryland, College Park (GPA: 3.85, graduating May 2026). He completed his B.Tech in Artificial Intelligence & Data Science at VVIT (Vasireddy Venkatadri Institute of Technology), Guntur, India (GPA: 3.6/4, 2020–2024). He builds agentic AI systems — usually solo — writing the prompts, the LangGraph routing, the FastAPI backend, and the React frontend himself.

## Contact

- Email: vikranthreddimasu@gmail.com
- GitHub: https://github.com/vikranthreddimasu
- LinkedIn: https://linkedin.com/in/vikranthreddimasu
- Website: https://vikranthreddimasu.xyz
- Resume: https://vikranthreddimasu.xyz/resume.pdf

## Projects

### Arya: Beating the OfficeQA Benchmark
Set the OfficeQA accuracy record (72.0%) in Sentient Labs Arena Cohort 0. Solo build — a single 6.3 KB Jinja2 system prompt, no MCP servers, no custom agent code. Runs at $0.07/task on the OpenHands-SDK harness with Minimax M2.5; peak score 191.0 (70.7% accuracy at $0.01/task) on the Goose harness.

- Tech: Python, Jinja2, OpenHands-SDK, Goose, Minimax M2.5, OpenRouter, Docker
- GitHub: https://github.com/vikranthreddimasu/arena-cli

### Pac-Man AI: From RL to World Models
PPO agent (4.2M params) learns Pac-Man across 128 parallel envs with curriculum learning. An RSSM world model (28M params) learns to dream the game, then a dream agent learns to play entirely from imagined gameplay.

- Tech: Python, PyTorch, PPO, RSSM, NumPy, MPS
- GitHub: https://github.com/vikranthreddimasu/pacman-ai

### WealthAgent: Multi-Agent AI for Financial Analytics
Multi-agent AI system using LangGraph and Anthropic Claude, routing financial queries to 3 specialized agents for analytics and TCA. Built with FastAPI, React, PostgreSQL, Docker. 85%+ test coverage.

- Tech: Python, LangGraph, Anthropic Claude, FastAPI, React, PostgreSQL, Docker
- GitHub: https://github.com/vikranthreddimasu/wealthagent

### MacFleet: Distributed ML for Apple Silicon
Pool Apple Silicon Macs into a distributed ML training cluster. Zero-config mDNS discovery, adaptive gradient compression (up to 200x over WiFi), thermal-aware scheduling, and dual engine support (PyTorch + MLX).

- Tech: Python, PyTorch, MLX, gRPC, mDNS, Ring AllReduce
- GitHub: https://github.com/vikranthreddimasu/MacFleet

### Offline Notebook LM
Offline-first RAG app for querying your own documents without cloud dependency. Supports 7+ file types, ~366 chunks/sec throughput.

- Tech: Python, Electron, React, FastAPI, ChromaDB, sentence-transformers
- GitHub: https://github.com/vikranthreddimasu/notebook-lm

### GAN for MNIST Synthesis
GAN trained on MNIST and deployed as a live demo on HuggingFace Spaces with CI/CD.

- Tech: Python, PyTorch, Gradio, HuggingFace Spaces, Docker
- GitHub: https://github.com/vikranthreddimasu/mnist-gan
- Live Demo: https://huggingface.co/spaces/rvikranth10/mnist-gan

### Distributed Transformer Training with Horovod
Custom GPT-style transformer trained from scratch on 34GB of BookCorpus across 4x H100 GPUs using Horovod on UMD's Zaratan HPC cluster.

- Tech: Python, PyTorch, Horovod, NCCL, SLURM, HuggingFace
- GitHub: https://github.com/vikranthreddimasu/umd_classes/tree/UmdTask31_Fall2025_Horovod_Distributed_Training_of_a_Transformer_Model_for_Text_Generation

## Skills

### AI & ML
Generative AI, Agentic AI, LLMs (Claude, OpenAI, Llama, Minimax), RAG, Text2SQL, LLM Fine-tuning & Distillation, Prompt Engineering, Benchmark Optimization, Distributed Training (DDP, DeepSpeed), Reinforcement Learning (PPO, World Models), Transformers, NLP

### Frameworks
LangGraph, LangChain, LlamaIndex, OpenHands-SDK, Goose, PyTorch, TensorFlow, FastAPI, React, Pydantic, WebSockets

### Languages
Python, SQL, R, TypeScript, Bash

### Databases & Cloud
PostgreSQL, ChromaDB, PGVector, Docker, CI/CD, GitHub Actions, HuggingFace Spaces

### Data & Visualization
Pandas, NumPy, Plotly, Streamlit, Jupyter, yfinance

## Experience

### Founder — Zero Sols (India, Aug 2023 – Jul 2024)
Founded and ran a software studio through college, growing it to a team of 6. Set vision and technical direction, owned client delivery, and shipped 20+ production applications for paying clients.

### AI/ML Intern — Amazon Web Services (AWS) (Hyderabad, India, Dec 2022 – Mar 2023)
Built and evaluated ML models for internal data-driven projects. Ran data cleaning, preprocessing, and EDA, supported feature engineering, training, and validation, and contributed to testing, monitoring, and cloud deployment.

## Beyond the Code (Leadership & Activities)

- Stanford University Innovation Fellow, Silicon Valley Meet-up at Stanford (March 2023): https://vikranthreddimasu.xyz/activities/stanford-meetup
- Change Forward Journal, Editorial Board Member, authored two published articles: https://vikranthreddimasu.xyz/activities/change-forward-journal
- Design Thinking Workshops, reached 5,000+ students across multiple universities: https://vikranthreddimasu.xyz/activities/design-thinking-workshops
- Zero Sols, founded web development startup: https://vikranthreddimasu.xyz/activities/zero-sols

## Resources & Methodologies

- The Golden Circle, Simon Sinek's framework for starting with 'Why': https://vikranthreddimasu.xyz/resources/golden-circle
- Design Thinking, human-centered approach to problem-solving: https://vikranthreddimasu.xyz/resources/design-thinking

## Status

Currently pursuing M.S. in Data Science (GPA: 3.85) at University of Maryland. Open to AI Engineer, ML Engineer, and Data Scientist roles.

## Links

- Extended profile: https://vikranthreddimasu.xyz/llms-full.txt
- All projects: https://vikranthreddimasu.xyz/projects
- Beyond the Code: https://vikranthreddimasu.xyz/activities
- Resources: https://vikranthreddimasu.xyz/resources