Sahil Garg

Builds and ships agentic, retrieval-augmented multi-agent LLM systems that integrate LangGraph/LangChain orchestration, vector search (Pinecone/FAISS), and robust validation and fallback mechanisms to enable accurate, low-latency QA and automation for finance and enterprise use cases. twine

Delivers production-grade AI services with strong MLOps and systems engineering: FastAPI microservices, Dockerized deployments, ONNX-optimized inference, Kafka streaming, AWS S3 and MongoDB storage, and targeted latency optimizations that reduced production response times by up to 2.2×. twine

Applies deep learning across computer vision, reinforcement learning, and multimodal models — from YOLO-based detection and OCR pipelines (license plates, solar defects) to TD3-based autonomous navigation and multimodal emotion recognition combining DistilBERT, DeepFace, and PANNs. twine

Leads small teams and cross-functional projects to translate research-quality models into stakeholder-facing products; holds a B.Tech in Computer Science with Data Science specialization and Cybersecurity honors (CGPA 8.3/10), and consistently demonstrates measurable impact (e.g., 97% mAP, 92% emotion accuracy, 70% reduction in manual BOQ extraction). scribd