About me
I'm a Master's student in Computer Science (AI/ML) at USC, graduating December 2025. I work at the intersection of ML systems and backend engineering — building custom CUDA/Triton kernels, LLM training and inference pipelines, and full-stack agentic AI applications.
My recent work spans vision-language model training, reinforcement learning for retrieval-augmented QA, and production-grade AI agents deployed on GCP. Previously, I built distributed event-streaming systems at Avaya processing 1M+ messages/day and conducted ML research on traffic prediction at CDAC. I'm passionate about making AI systems that are fast, reliable, and actually useful.
My education & experience
Master's in Computer Science (AI/ML) @ University of Southern California
Los Angeles, United States
Specialization in AI/ML. Coursework: Machine Learning (CSCI 567), Artificial Intelligence (CSCI 561), Deep Learning, Natural Language Processing, Distributed Systems.
01/2024 - 12/2025Graduate Research Assistant @ USC
Los Angeles, United States
Built real-time WebSocket event-driven backend systems achieving sub-100ms synchronization latency. Designed scalable architectures for concurrent data processing pipelines.
08/2024 - 12/2024Software Development Engineer Intern @ Avaya
Pune, India
Worked on Avaya Social Connections, a cloud-based CCaaS platform integrating contact centers with Facebook, Instagram, WhatsApp, and Twitter. Built microservices using Java, Spring Boot, Node.js, and TypeScript. Implemented Kafka-based event streaming processing 1M+ messages/day. Added Datadog monitoring, PII hashing, graceful shutdown. Increased test coverage from 30% to 80%.
01/2022 - 06/2022ML Research Intern @ CDAC (Centre for Development of Advanced Computing)
Pune, India
Developed LSTM/GRU models for traffic matrix prediction on telecom datasets (Abilene, GÉANT). Built Docker-based CI/CD pipeline for model deployment. Discovered that predicting overall traffic matrix with key element correction optimally balances performance and prediction time.
07/2022 - 06/2023B.E. in Computer Science @ SRM Institute of Science and Technology
Chennai, India
Major in Computer Science with specialization in Big Data Analytics. GPA: 9.3/10.
08/2018 - 04/2022My projects
Vision-Language Model (VLM) Training
Trained a custom VLM combining SigLIP vision encoder, MLP projector, and Qwen2.5-0.5B language model on LLaVA-Instruct-150K dataset. Implemented KV-cache optimization, INT4/INT8 quantization, continuous batching, and custom Triton kernel fusion for...
- PyTorch
- Triton
- CUDA
- Qwen2.5
- SigLIP
- Quantization
Write-in-Margins (WiM) RL System
Retrieval-augmented QA system for HotpotQA where an LLM generates margin notes on document chunks before synthesizing answers. Trained with PPO on Qwen2.5-3B using NF4 quantization, LoRA/PEFT, and PagedAdamW8bit. Extended the...
- PPO
- LoRA
- Qwen2.5
- RAG
- HuggingFace
- W&B
VC Deal Sourcing Agent
LangGraph-based agentic system for venture capital deal sourcing with Anthropic tool-calling, Pydantic v2 anti-hallucination validators, and multi-source search router (Exa, Tavily, TechCrunch RSS, ProductHunt). Uses SqliteSaver persistence and source-weighted result...
- LangGraph
- Anthropic API
- Pydantic
- Exa
- Tavily
GraphRAG Knowledge Retrieval
Built a hybrid retrieval system using Neo4j knowledge graphs combined with vector similarity search and cross-encoder reranking. Converts e-commerce product reviews into structured graph representations enabling both graph traversal and...
- Neo4j
- LangChain
- RAG
- OpenAI
- Python
My skills
- Python
- C++
- CUDA
- Triton
- PyTorch
- TensorFlow
- LangGraph
- LangChain
- HuggingFace
- RAG
- LLM Fine-tuning
- Distributed Systems
- FastAPI
- Node.js
- TypeScript
- React
- Next.js
- PostgreSQL
- Neo4j
- Docker
- GCP
- AWS
- Kafka
- Git
- Tailwind CSS
Contact me
Please contact me directly at spgore@usc.edu