Mahesh Godike

AI Engineer with 1+ year of production experience building and deploying GenAI agents, RAG pipelines, and backend services

Talent

HyderabadMember since July 27, 2026

LangChain AgentsMulti-step ReasoningReflection-based ValidationRAG PipelinesOpenAI APIHugging Face TransformersPythonFastAPIChromaDBFAISSDockerGit/GitHubSegFormerDeepLabV3NVIDIA Jetson AGX Orin

Hire this person

Send a job offer directly to this candidate

LangChain AgentsMulti-step ReasoningReflection-based ValidationRAG PipelinesOpenAI API

About

AI Engineer with 1+ year of production experience building and deploying GenAI agents, RAG pipelines, and backend services for real business use cases. Hands-on with LangChain agent development (tool definition, multi-step reasoning, reflection-based validation), RAG architectures using ChromaDB/FAISS, and LLM integration via Hugging Face and OpenAI APIs. Experienced shipping RESTful APIs (FastAPI), containerized services (Docker), and version-controlled codebases (Git/GitHub) into production, including edge and enterprise environments.

Comfortable working cross-functionally and adapting quickly to new agent frameworks and tools.

Experience

AI Engineer (Promoted from Intern to Full-Time) - Valkontek Embedded & IoT Services Pvt. Ltd.

Hyderabad, India

(2025-05 - 2026-06)

Identified slow internal document search as a recurring cross-team bottleneck; scoped requirements and built an on-premise RAG pipeline (LangChain + ChromaDB/FAISS) over 500+ proprietary documents, containerized with Docker and version-controlled via Git, significantly cutting query resolution time.
Built an Agentic AI WorkTracker (LangChain agents) by defining tools, prompts, and multi-step reasoning workflows to monitor file-system, Git, and app-usage signals, with reflection-based checks on report accuracy – eliminating manual status reporting for 10+ remote developers.
Designed and deployed a real-time Person Re-Identification pipeline (ByteTrack + StrongSORT) as a FastAPI microservice, sustaining 24/7 surveillance across 8 camera feeds with identity continuity through occlusions and frame drops.
Engineered and deployed a multi-model automated vehicle inspection pipeline (YOLOv8 + DeepLabV3 + SegFormer) covering 7 defect classes on NVIDIA Jetson AGX Orin, achieving <50ms inference latency and 90%+ detection accuracy in production.
Improved model inference throughput by ~35% over baseline PyTorch deployment through quantization and batching optimization, strengthening system reliability and scalability on Jetson AGX Orin.