Applied AI Engineer - LLM Systems & Production AI

Samarth Yadannavar

Applied AI Engineer — LLM Systems & Production AI

*******************@*****.*** +91-703******* Bangalore, India linkedin.com/in/samarth-yadannavar samarthy06.github.io/Portfolio Seeking relocation to Europe Available for visa sponsorship SUMMARY

Applied AI Engineer with 1.5+ years of experience designing and building AI-driven solutions, LLM integrations, and scalable enterprise deployment pipelines in production. Delivered measurable outcomes for 10K+ users through real- time AI platforms (<300ms), agentic workflows (LangGraph), and CI/CD-driven deployment with observability. Strong software engineering fundamentals and communication skills with end-to-end ownership from prompt engineering to containerized deployment, contributing to engineering standards across cross-functional teams. WORK EXPERIENCE

QuantAI, Bangalore Sep 2024 – Present

AI Engineer — LLM Systems & Agentic AI (formerly AI Scientist)

• Engineered a production-grade Agent Assist platform with sub-300mslatency,integrating OpenAI and Claude LLM APIs via async execution and streaming for 200+ concurrent agents, driving a 12% CSAT increase (~$240K annual retention impact).

• Developed a reusable agentic project manager on LangGraph with custom tool-calling, sliding-window memory, and persistent state — automating 1K+ JIRA tasks/month and saving ~120 engineering hours monthly across 3 product teams.

• Designed an enterprise-scale knowledge-graph-augmented RAG system for 10K+ users, combining graph-based re- trieval with vector search (FAISS, pgvector) to improve contextual relevance by 28% and reduce cold-start failures by 35%.

• Fine-tuned domain LLMs (Mistral-7B, Llama-3) on 50K+ samples using SFT and DPO; deployed quantized SLMs via vLLM for sub-100msmodel serving in production pipelines, tracked with MLflow and Weights & Biases. Fractal Analytics, Mumbai Apr 2024 – Jul 2024

AI Research Intern — Generative AI & NLP

• Built ‘Med-Agents’, a multi-agent CoT reasoning framework integrating GPT-4o for medical diagnostics, improving accuracy from 77% to 83% on the NEET PG benchmark; applied DPO for LLaVA-Med to improve multimodal medical image interpretation.

PROJECTS

ProfessorOS — Multi-Agent Learning Platform Jan 2026 – Present GitHub

• Built a scalable multi-agent platform with LangGraph orchestrating 4 agents (Planner, Teacher, Quiz, Progress) over Temporal workflows and pgvector; deployed on AWS ECS with Docker, Prometheus/Grafana observability, and BYOK architecture — reducing infra costs by 30%.

LLM-as-a-Judge Evaluation Framework Nov 2025 – Jan 2026 GitHub

• Engineered an automated LLM evaluation framework (G-Eval) scoring 5K+ outputs across relevance, faithfulness, and hallucination metrics for 20+ agentic configurations; improved production agent performance by 22% via CI/CD- integrated prompt optimization.

SKILLS

Languages & Backend: Python, JavaScript, C++, SQL, FastAPI, GraphQL, Node.js LLM & Agentic AI: LangChain, LangGraph, LlamaIndex, OpenAI API, Claude API, Hugging Face, vLLM, Prompt Engineering, ReAct, Tool-Calling, RAG, Hybrid Search, Vector DBs (FAISS, Pinecone, pgvector), Knowledge Graphs, NeMo Guardrails, Responsible AI Model Development: Fine-Tuning (SFT, DPO, RLHF), Quantization (GGUF, GPTQ), LLM Evaluation (G-Eval, LLM-as-a-Judge)

Infrastructure & CI/CD: Docker, Kubernetes, CI/CD, Terraform, AWS (SageMaker, Bedrock, ECS), Prometheus, Grafana

MLOps & Observability: MLflow, Weights & Biases, Evidently AI, Experiment Tracking, Model Governance EDUCATION

Indian Institute of Technology (IIT) Bombay Jul 2020 – May 2024 B.Tech — Metallurgical Eng. & Materials Science Minor: Data Science & AI PUBLICATIONS

“Multimodal Analysis of Learning-Centered Emotions and Cognitive Processes in Open-Ended Learning Environments”

— IEEE, 2024

Contact this candidate