Arnav Panigrahi
Software Developer, AI/ML Enthusiast
Building ethical AI systems that solve real-world problems
About Me
I am a software engineer with an MS in Computer Science from UC Riverside, specializing in building robust full-stack applications and intelligent AI/ML systems. With over two years of professional experience, I excel at developing software from concept to deployment.
In my current role at AllCheer, I engineered a core HIPAA-compliant enterprise system using AI and modern cloud infrastructure. This initiative resulted in $20K in annual savings and a 60% reduction in processing time.
I'm driven by a passion for building responsible technology, with a focus on ethical AI, user privacy, and the principles of open-source development.
My Skills
Languages & Core Technologies
- Python
- JavaScript
- TypeScript
- HTML5
- CSS3
- SQL
- JSON
- YAML
AI & Machine Learning
- LangChain
- LangGraph
- LLMs
- RAG
- PyTorch
- TensorFlow
- Scikit-learn
- NumPy
- Pandas
Backend Development
- FastAPI
- Flask
- Node.js
- Express.js
- RESTful APIs
Frontend Development
- React
- Astro
- Angular
- Next.js
- Tailwind CSS
- WCAG Accessibility
Cloud & DevOps
- AWS
- GCP
- Azure
- Docker
- Kubernetes
- CI/CD
- Git
Databases & Vector Stores
- PostgreSQL
- MongoDB
- MySQL
- Redis
- Pinecone
My Projects
bedtime.ai – Multi-Modal AI Storytelling Platform
Architected a multi-modal AI pipeline that transforms children's drawings into personalized stories narrated in a cloned parent's voice. Engineered the full-stack solution, integrating a fine-tuned phi-3 LLM with EfficientNet for image analysis and coqui-ai for voice synthesis. The system is containerized with Docker and delivers complete stories in under 30 seconds.
Abliteration: LLM Uncensoring via Weight Modification
Implemented abliteration techniques to remove censorship from LiquidAI/LFM2.5-1.2B-Instruct by identifying and orthogonalizing refusal directions in residual stream activations. Developed inference-time intervention and direct weight modification approaches, successfully uncensoring content while preserving model capabilities.
Agentic Job Tracker & Helper
Engineered a stateful AI agent using LangGraph to automate the entire job application workflow. The agent uses natural language to manage application statuses in a Supabase database, parse job descriptions, and automatically tailor a user's resume by generating targeted modifications.
Rethink BH Automation – FastAPI + Cloud Run Backend
Built a production FastAPI service to sync real-time appointment and authorization data from Rethink BH to Supabase by emulating secure user login. Achieved 90% cost savings and reduced sync latency from 24-48h to real-time. Deployed with Docker, Secret Manager, and full observability stack on Google Cloud Platform.
RAG for Document Intelligence
Architected a high-throughput RAG system for semantic search over PDF datasets. The system leverages Pinecone for vector storage and Groq for accelerated inference, achieving a 35% improvement in search relevance over traditional keyword-based methods while maintaining sub-second response times. Implemented custom data chunking strategies to optimize retrieval accuracy.
Enterprise ROI System with AI Agent
Led the re-engineering of a manual Release of Information (ROI) workflow into a fully automated, AI-driven system. The AI agent I developed manages the entire document lifecycle with enterprise-grade security, directly resulting in a $20K annual cost saving and a 60% reduction in processing time while adhering to HIPAA compliance.