Unlike AI researchers who only know Python scripts, Dherk spent years in commercial data centers keeping real Linux servers, firewalls, and networks running 24/7 without crashing (99.9% uptime).
He fine-tunes custom AI models (like handwriting readers and domain assistants) and optimizes their serving engines so they respond in milliseconds and serve dozens of users simultaneously on less hardware.
He builds teams of AI "agents" that work together right on your own desktop without sending your private company documents or personal chats to external clouds.
dherk > orchestrating deep neural runtimes.
System Engineer transitioning to AI Engineering, currently pursuing M.Tech in AI & Data Science at GEC Barton Hill. Hands-on experience fine-tuning OCR models (Microsoft TrOCR-large at 93.3% word accuracy on Hugging Face), orchestrating high-throughput LLM workloads (vLLM, SGLang), constructing local multi-agent swarms, and architecting production-hardened RAG topologies.
Interactive Skill Map: From Infrastructure to AI Systems
Explore how hands-on server engineering, deep learning training, and high-speed AI inference connect together. Hover or tap any node to see what it does in simple terms.
vLLM & SGLang Serving
PagedAttention & Continuous Batching
High-concurrency LLM serving pipelines leveraging PagedAttention, KV cache management, continuous batching, and tensor-parallel deployment.
High-speed engine that lets AI language models answer dozens of people at the exact same time without crashing or slowing down, cutting wait times by 42%.
16-Domain Competency Matrix & Interactive Guide
Complete catalog of programming languages, deep learning frameworks, hosting panels, and systems infrastructure with clear, practical summaries.
1. Programming Languages
Core DevelopmentThe core coding languages used to build high-performance backend servers, system scripts, database queries, and deep learning algorithms.
2. ML & DL Frameworks
Neural Network EnginesThe mathematical software engines used to train neural networks, analyze data, and build web APIs that power artificial intelligence.
3. LLM & NLP Tools
Inference & Fine-TuningCutting-edge tools for fine-tuning language and vision models, speeding up chatbot responses, compressing models, and building smart question-answering systems.
4. Training & Evaluation
Distributed OptimizationTechniques for training AI models using multiple graphics cards in parallel, teaching models custom alphabets (like cursive handwriting), and scientifically verifying accuracy.
5. Vector Databases
Semantic IndexingIntelligent databases that store information by conceptual meaning rather than exact words, enabling AI to instantly find relevant facts without hallucinations.
6. Cloud & Platforms
Certified ArchitectureEnterprise cloud hosting platforms that provide powerful on-demand computing power, secure networks, and automated software deployment.
7. Operating Systems
Enterprise FleetHeavy-duty server operating systems configured for 24/7 reliability, running mission-critical enterprise applications without downtime.
8. Virtualization & Containers
Hypervisors & SandboxesTechnologies that carve physical servers into isolated virtual machines or containers, maximizing hardware efficiency and security.
9. Web & Server Stack
LAMP EnvironmentsClassic, battle-tested web hosting stacks capable of serving high-traffic websites, processing customer database records, and managing web code.
10. DevOps & Automation
Monitoring & BotsAutomated watchdogs that monitor server health 24/7, alert on problems instantly, and internal AI chatbots that automate daily employee tasks.
11. Hosting & DNS
Panel MigrationsManagement of internet domain names, email routing, and complex zero-downtime website transfers between hosting providers.
12. Security & Hardening
Auditing & BackupsFortifying servers against cyberattacks, setting up encrypted connections (HTTPS), running automated safety backups, and testing for security flaws.
13. Storage Solutions
Private Cloud FilesPrivate, on-premise cloud storage systems that give organizations full ownership of their files without relying on commercial third parties.
14. Version Control
Git OperationsProfessional code versioning and collaboration platforms enabling safe code review, history tracking, and automated deployment pipelines.
15. Technical Leadership
Strategy & DevelopmentCommunicating complex engineering concepts clearly, guiding project direction, solving ambiguous challenges, and partnering smoothly across teams.
16. Mentorship & CSR
Team UpskillingRunning technical training sessions for junior engineers, coordinating cross-department initiatives, and volunteering skills for educational community outreach.
Experience & Systems Evolution
AI / ML Intern — CDIT
Centre for Development of Imaging Technology
- Fine-tuned Microsoft TrOCR-large on handwritten Malayalam datasets achieving 93.3% word accuracy and 1.1% CER.
- Published model artifacts to Hugging Face Model Hub with interactive demo space.
- Orchestrated multi-GPU distributed fine-tuning runs on AWS compute instances with automated checkpoint telemetry.
- Engineered high-concurrency model serving backends leveraging vLLM and SGLang.
System Engineer — HostDime India Pvt Ltd
Global Edge Data Center Operations
Detail-oriented System Administrator with hands-on experience in managing Linux/Windows servers, virtualization platforms, and automation using Bash scripting. Skilled in deploying open-source solutions, optimizing server infrastructure, and leading internal tech initiatives that drive operational efficiency and team productivity.
Featured AI Systems & Deployments
Turi — Local Multi-Agent Assistant
Privacy-first local orchestration mesh synchronizing 5 specialized models (Gemma 4, Phi-4 Mini, Qwen 2.5, Llama 3.2). 3-tier memory graph, Whisper STT, Piper TTS, and custom MCP tools.
Patient Safety Sentinel (PSS)
Automated adverse drug reaction crawler monitoring Reddit, Twitter, and medical RSS feeds. Thread-safe ingestion workers, APScheduler cron scheduling, and SQLite deduplication.
TrOCR Malayalam OCR Model
Fine-tuned Vision-Encoder-Decoder model customized with custom character tokenizers for curved Indic Malayalam scripts. Benchmarked against synthetic and cursive scripts.
Educational Foundations & Certifications
M.Tech in AI & Data Science
Govt. Engineering College Barton Hill (GEC Barton Hill)
B.Tech in Electronics & Communication
University College of Engineering (UCE), Kariyavattom
Human Optimization Protocol
Let's Connect & Collaborate
Whether you are hiring for AI or Systems engineering, want to collaborate on high-speed inference and vision OCR, or just want to say hi, feel free to reach out directly through email or LinkedIn!
Email Directly
Send a message straight to my inbox. Click below to open your email client with a pre-composed introductory note.
Connect on LinkedIn
View my professional timeline, recommendations, colleagues, and send direct in-app messages on LinkedIn.
Repositories & Models
Inspect public GitHub source code, star repositories, or download vision OCR model weights and datasets.