# Sohail Gidwani

> AI Engineer and Full-Stack Developer based in Los Angeles, CA.
> M.S. Computer Science at USC. Specializing in RAG systems, LLM integration, and applied AI.

## Site

- Homepage: https://sohailgidwani.app
- Projects: https://sohailgidwani.app/projects
- Resume (PDF): https://sohailgidwani.app/documents/Sohail_Gidwani_Resume.pdf
- Resume (JSON): https://sohailgidwani.app/resume.json
- MCP Server: https://sohailgidwani.app/api/mcp (POST JSON-RPC 2.0 — resources: portfolio://profile, portfolio://projects, portfolio://experience, portfolio://education)

## About

Sohail Gidwani is an AI/ML engineer and full-stack developer graduating with a Master of Science in Computer Science from the University of Southern California (Viterbi School of Engineering) in May 2027. He builds production AI systems — RAG pipelines, multi-modal deep learning, LLM-powered applications — and pairs them with clean, user-facing frontends. Previously a Full-Stack Software Developer at IIFL Finance Ltd and Senior Software Engineer at Insaito, Inc.

## Current Role

Research Assistant at Keck School of Medicine of USC, working on multi-modal deep learning for Alzheimer's disease prediction using neuroimaging and clinical data across 71M+ parameter models.

## Key Skills

- AI/ML: PyTorch, TensorFlow, Scikit-learn, HuggingFace, LangChain, OpenCV, RAG, LLMOps, MCP
- Languages: Python, TypeScript, JavaScript, Java, C++, SQL
- Web: React, Next.js, Node.js, Flask, FastAPI, Hono, Express, Tailwind CSS
- Databases: PostgreSQL, MongoDB, Qdrant, Redis, pgvector, MySQL, Elasticsearch
- Cloud: AWS, Azure, Google Cloud, Cloudflare, Docker, Kubernetes
- Methods: Agile, TDD, CI/CD, RESTful APIs, Serverless

## Research

- Multi-Modal VLM + RAG VQA for Alzheimer's Detection: Keck USC — 2,363 ADNI subjects, 0.707 balanced accuracy, missing-modality cross-attention fusion, RAG VQA benchmarking Mistral 7B vs Gemma 4 26B vs MedGemma 1.5 4B (https://sohailgidwani.app/research/multimodal-alzheimers-vqa)
- CoT Faithfulness Analysis: USC CSCI-544 NLP study — 4 experiments (SCR, CFR, SBH) probing chain-of-thought faithfulness in Llama 3.2 3B and Qwen 2.5 7B across GSM8K and ARC-Challenge; ~15K queries via Ollama (https://sohailgidwani.app/projects/cot-faithfulness)

## Projects

- Knowledge Hub: Document management with OCR, semantic search (pgvector), and RAG Q&A via Ollama (https://sohailgidwani.app/projects/knowledge-hub) — Technical Deep Dive: https://sohailgidwani.app/projects/knowledge-hub/deep-dive
- Image Captioning: CNN + Transformer pipeline achieving 0.80 BLEU, deployed via Streamlit (https://sohailgidwani.app/projects/image-captioning)
- ScribeGlobe: Medium-style blogging platform on Cloudflare Workers with Hono + React (https://sohailgidwani.app/projects/scribeglobe)
- Tech Updates: AI-powered news aggregator using Azure OpenAI and Qdrant for semantic deduplication (https://sohailgidwani.app/projects/tech-updates)

## Education

- M.S. Computer Science, University of Southern California (2025–2027), GPA 3.5/4.0
- B.E. Computer Engineering, University of Mumbai - TSEC (2019–2023), CGPA 9.05/10

## Contact

- Email: sohailgidwani15@gmail.com
- LinkedIn: https://linkedin.com/in/sohail-gidwani/
- GitHub: https://github.com/SohailGidwani
- Portfolio: https://sohailgidwani.app

## Machine-Readable Data

- JSON Resume: https://sohailgidwani.app/resume.json (jsonresume.org v1 schema)
- MCP Server: https://sohailgidwani.app/api/mcp (MCP protocol 2024-11-05, POST JSON-RPC 2.0)
- LLMs.txt (full): https://sohailgidwani.app/llms-full.txt
- LLMs.txt (well-known): https://sohailgidwani.app/.well-known/llms.txt

## Detailed Version

For a more comprehensive version of this file, see: https://sohailgidwani.app/llms-full.txt