{ "$schema": "https://raw.githubusercontent.com/jsonresume/resume-schema/v1.0.0/schema.json", "basics": { "name": "Sohail Gidwani", "label": "AI/ML Engineer & Full-Stack Developer", "image": "https://sohailgidwani.app/api/og", "email": "sohailgidwani15@gmail.com", "url": "https://sohailgidwani.app", "summary": "AI/ML engineer and full-stack developer graduating with an M.S. in Computer Science from USC in May 2027. Builds production AI systems — RAG pipelines, multi-modal deep learning, LLM-powered applications — and pairs them with clean, user-facing frontends.", "location": { "city": "Los Angeles", "region": "CA", "countryCode": "US" }, "profiles": [ { "network": "GitHub", "username": "SohailGidwani", "url": "https://github.com/SohailGidwani" }, { "network": "LinkedIn", "username": "sohail-gidwani", "url": "https://linkedin.com/in/sohail-gidwani/" } ] }, "work": [ { "name": "Keck School of Medicine of USC", "position": "Student Worker - Research Assistant", "url": "https://keck.usc.edu", "startDate": "2025-10-01", "summary": "Multimodal deep learning for Alzheimer's disease classification and VQA using T1 MRI, DTI imaging, and clinical data.", "highlights": [ "Architected multimodal pipeline (MEMOIR-VLM) across 2,363 ADNI subjects; 70.7% balanced accuracy on 3-class diagnosis, 93.3% on binary (CN vs Dementia)", "Designed missing-modality fusion via cross-attention with stochastic modality dropout; robust inference with any subset of T1, DTI, and clinical inputs (39.4% DTI coverage)", "Built end-to-end experimentation infrastructure: two-stage training (CLIP contrastive pre-training → multi-task fine-tuning), modality ablation across 7 combinations, ~70M parameter models" ] }, { "name": "Insaito, Inc.", "position": "Senior Software Engineer - I", "startDate": "2025-05-01", "endDate": "2025-07-31", "summary": "Led architecture of an AI agent builder platform enabling custom workflow creation with OAuth integrations.", "highlights": [ "Led architecture of AI agent builder platform with OAuth integrations for 100+ third-party apps", "Built core infrastructure for open-source LLM deployment (Qwen 3, Mistral Small 24B) and Model Context Protocol (MCP) servers", "Designed serverless backend for agent orchestration supporting concurrent multi-step workflows with tool-calling" ] }, { "name": "IIFL Finance Ltd.", "position": "Full Stack Software Developer", "url": "https://iifl.com", "startDate": "2023-06-01", "endDate": "2025-05-31", "summary": "Built AI-powered internal tools and financial applications for a leading NBFC.", "highlights": [ "Built internal RAG chatbot using Python, Flask, Qdrant, and Azure OpenAI; integrated with Zoho ticketing to automate employee support", "Engineered AI-powered Gold Loan Image Audit system using GroundingDINO and Swin-Transformer; reduced potential loan fraud by 15%", "Designed CapitalGenie, an automated support system with GPT-4o that accelerated issue resolution time by 70%" ] } ], "education": [ { "institution": "University of Southern California", "url": "https://www.usc.edu", "area": "Computer Science", "studyType": "Master of Science", "startDate": "2025-08-01", "endDate": "2027-05-31", "score": "3.5/4.0", "courses": [ "Analysis of Algorithms", "Information Retrieval & Web Search Engines", "Natural Language Processing", "Machine Learning for Data Science" ] }, { "institution": "Thadomal Shahani Engineering College, University of Mumbai", "url": "https://tsec.edu", "area": "Computer Engineering", "studyType": "Bachelor of Engineering", "startDate": "2019-08-01", "endDate": "2023-05-31", "score": "9.05/10", "courses": [ "Artificial Intelligence", "Machine Learning", "Advanced DBMS", "Data Structures & Algorithms", "Big Data Analytics", "Cloud Computing" ] } ], "skills": [ { "name": "AI / Machine Learning", "keywords": ["PyTorch", "TensorFlow", "Keras", "Scikit-learn", "HuggingFace", "LangChain", "RAG Systems", "LLMOps", "MCP", "OpenCV", "NLTK", "spaCy", "Vector Databases"] }, { "name": "Programming Languages", "keywords": ["Python", "TypeScript", "JavaScript", "Java", "C++", "SQL"] }, { "name": "Web Development", "keywords": ["React", "Next.js", "Node.js", "Flask", "FastAPI", "Hono", "Express", "Tailwind CSS"] }, { "name": "Databases", "keywords": ["PostgreSQL", "MongoDB", "Qdrant", "Redis", "pgvector", "MySQL", "Elasticsearch"] }, { "name": "Cloud & DevOps", "keywords": ["AWS", "Azure", "Google Cloud", "Cloudflare", "Docker", "Kubernetes", "CI/CD"] } ], "projects": [ { "name": "Knowledge Hub", "description": "Document management with OCR, semantic search (pgvector), and RAG Q&A via local LLM (Ollama). Built for personal USC coursework management.", "highlights": [ "OCR pipeline for handwritten notes and PDFs using Tesseract", "pgvector semantic search alongside full-text search", "RAG Q&A with Ollama (gemma3:1b) citing source passages", "Containerized with Docker — setup is docker-compose up" ], "keywords": ["Flask", "PostgreSQL", "pgvector", "RAG", "OCR", "Ollama", "Python"], "url": "https://sohailgidwani.app/projects/knowledge-hub", "roles": ["Creator"], "type": "application" }, { "name": "CoT Faithfulness Analysis", "description": "NLP research project probing whether chain-of-thought reasoning in LLMs causally drives answers or is post-hoc rationalization.", "highlights": [ "4 experiments: SCR (step truncation), CFR (corruption), SBH (biased hints)", "Models: Llama 3.2 3B and Qwen 2.5 7B via Ollama", "Benchmarks: GSM8K (math) and ARC-Challenge (science)", "~15K queries total; finds CoT is partially causal, partially decorative" ], "keywords": ["Python", "LLM", "Ollama", "NLP", "Research", "Llama", "Qwen"], "url": "https://sohailgidwani.app/projects/cot-faithfulness", "roles": ["Researcher"], "type": "research" }, { "name": "Image Captioning", "description": "CNN + Transformer pipeline for automatic image caption generation, achieving 0.80 BLEU score.", "highlights": [ "VGG-16 feature extraction + LSTM and Transformer caption generators", "Transformer model: 0.80 BLEU vs 0.65 LSTM baseline (+23%)", "Deployed as Streamlit web app for live demo" ], "keywords": ["TensorFlow", "CNN", "Transformer", "VGG-16", "BLEU", "Streamlit", "Python"], "url": "https://sohailgidwani.app/projects/image-captioning", "roles": ["Creator"], "type": "application" }, { "name": "ScribeGlobe", "description": "Medium-style blogging platform built on Cloudflare Workers edge infrastructure.", "highlights": [ "React + Vite frontend with full auth flow", "Hono API running on Cloudflare Workers (edge)", "PostgreSQL (Neon) for persistent storage" ], "keywords": ["React", "TypeScript", "Cloudflare Workers", "Hono", "PostgreSQL", "Neon"], "url": "https://sohailgidwani.app/projects/scribeglobe", "roles": ["Creator"], "type": "application" }, { "name": "Tech Updates", "description": "AI-powered news aggregator with semantic deduplication using Azure OpenAI and Qdrant.", "highlights": [ "Automated news ingestion pipeline from multiple RSS feeds", "Qdrant vector DB for semantic duplicate detection", "Azure OpenAI for article summarization and tagging" ], "keywords": ["Python", "Flask", "Azure OpenAI", "Qdrant", "SentenceTransformers", "React"], "url": "https://sohailgidwani.app/projects/tech-updates", "roles": ["Creator"], "type": "application" } ], "languages": [ { "language": "English", "fluency": "Fluent" }, { "language": "Hindi", "fluency": "Native speaker" }, { "language": "Sindhi", "fluency": "Native speaker" } ], "meta": { "canonical": "https://sohailgidwani.app/resume.json", "version": "v1.0.0", "lastModified": "2026-04-26" } }