AI Engineer | Full Stack Developer | System Design
Full-stack developer and AI engineer with 2.5+ years of experience building production AI systems, microservices platforms, and scalable infrastructure. Strong system design fundamentals -- experienced in architecting distributed systems, agentic workflows, and data-intensive applications from the ground up.
Specialized in integrating LLMs, computer vision, and speech models into real-world systems with robust deployment pipelines. Proven track record across industrial operations, mining safety, and government sectors.
- Currently building AI-powered tools at VyomChara -- a drone tech and AI company based in India
- Work spans RAG systems, agentic AI workflows, multimodal LLM integration, CV pipelines, and Kubernetes-managed ML infrastructure
- Always learning: distributed systems, AI engineering, and scalable architecture patterns
- Ask me about web dev, AI/ML, system design, or turning research into working products
SDE | Jan 2024 -- Present | Hybrid
- Architected an AI-powered case management platform with agentic investigation workflows -- autonomous task delegation across specialized agents for document processing, evidence analysis, and legal vetting
- Developed an on-premise RAG system combining vector database retrieval with full-text indexing for real-time text/image Q&A
- Built computer vision pipeline APIs for real-time vehicle/personnel/hazard detection in mining operations
- Set up and managed a private AI data center with Kubernetes clusters for ML workloads
- Designed backend systems supporting drone telemetry and low-latency video streaming over 5G networks
- Engineered company website achieving 90+ Lighthouse score with 99.9% uptime supporting 3K+ monthly visitors
- Built deployment pipelines and dev environments for startups incubated under the TEXMiN foundation
Software Engineering Intern | May 2025 -- Jul 2025
- Benchmarked multiple deep learning architectures (3D CNN, HybridSN, Capsule Networks) for mineral segmentation, achieving 99%+ validation accuracy
- Architected microservices backend for hyperspectral image analysis; containerized 5 specialized services with GPU detection and memory management
- Operationalized ML models via REST APIs supporting high-throughput batch processing for industrial use cases
- Built React frontend with responsive scientific data visualization interfaces
Project Intern | Aug 2023 -- Jan 2024
- Analyzed 10+ academic papers on hyperspectral mineral classification, benchmarking model performance on EO1-Hyperion datasets
- Authored comprehensive comparison report of deep learning models for mineral classification
- Mastered hyperspectral camera operation, capturing and processing high-resolution spectral data from 7+ mineral types
Most of my work is in private repositories due to client and company confidentiality. Here's a snapshot of what's behind closed doors:
- AI Case Management Platform -- Full-stack platform with agentic investigation workflows, multi-role RBAC, SSE streaming, and pluggable LLM service layer supporting cloud and local providers
- On-Premise RAG System -- Vector database + full-text search hybrid with multimodal analysis, web search grounding, and self-correcting response generation
- 3D AI Avatar Interface -- Real-time animated avatar with speech-to-text, text-to-speech, and lip-sync integration; optimized GPU memory for large models
- Computer Vision Pipelines -- Real-time vehicle/personnel/hazard detection APIs for mining operations with object storage and database integration
- Hyperspectral Analysis Platform -- Microservices architecture with 5 containerized services, GPU detection, and 99%+ validation accuracy on mineral segmentation
- Private AI Data Center -- Kubernetes-managed ML infrastructure with deployment pipelines for multiple startup environments
- Drone Telemetry Systems -- Backend systems for real-time telemetry and low-latency video streaming over 5G networks
If you'd like to discuss my work in detail or see code samples, feel free to reach out. I'm happy to walk through architecture decisions, code samples, and system designs in interviews.
- Email: talktonaman@duck.com
- LinkedIn: linkedin.com/in/kumar-naman-tiwary-b16621214
- Resume available upon request