Dinesh
Solanki
AI Specialist & Generative AI Engineer building intelligent systems at the frontier of LLMs, RAG, and autonomous agents. Samsung R&D alumnus · Springer Published Researcher.
About Me
My journey into AI began during my B.Tech in Artificial Intelligence & Machine Learning at Cambridge Institute of Technology, Bengaluru, where I shifted focus from theory to shipping real systems. By final year, I had published a peer-reviewed paper at Springer Nature Singapore (SmartCom 2025) on Generative AI-powered video summarization using CNN-Transformer architectures.
That foundation led to Samsung R&D Institute India, where I progressed from Research Intern to Team Lead. I built an AI Video Q&A chatbot (Video-LLaVA-7B + Whisper ASR, 96% transcript accuracy), then led the engineering of a platform-agnostic Stable Diffusion pipeline - bypassing AUTOMATIC1111 UI dependencies to run SDXL, ControlNet, and IP-Adapter workflows entirely from CLI. That work earned the Outstanding and Excellence Awards from SRI-B.
At Trilliant Digital, I applied those skills to production agentic systems: multi-agent pipelines with CrewAI and LangChain, RAG architectures over Supabase vector stores, and N8N workflow orchestrations that cut manual marketing effort by 60% and tripled content delivery speed.
Today I focus on building AI systems that span the full stack: from LLM fine-tuning (LoRA, DreamBooth) and RAG pipeline design to Dockerized cloud deployments on GCP, AWS SageMaker, and Azure ML. Every project I take on has one goal: measurable, production-grade impact.
Education
Cambridge Institute of Technology
Bengaluru, Karnataka, India
Key Coursework & Specializations
Graduated with a strong foundation in AI/ML theory and practical implementation. Published peer-reviewed research at SmartCom 2025 (Springer Singapore) on Generative AI-Powered Video Summarization. Conducted the Generative AI Workshop at the World Resources Webinar, covering GANs, VAEs, Transformers, and ethical AI.
Publications
Peer-reviewed contributions advancing state-of-the-art in generative AI and multimedia intelligence.
Generative AI-Powered Tool for Automated Video Summarization
Proceedings of SmartCom 2025, Volume 4
This research introduces a generative AI framework implementing hierarchical spatio-temporal video content analysis through integrated CNNs for multi-scale feature extraction, transformer-based LLMs for semantic understanding, and multimodal NLP techniques with sophisticated encoder-decoder architectures and attention mechanisms.
Generative AI Workshop
Presented at the World Resources Webinar
Delivered a comprehensive session on Generative AI fundamentals, architectures, and real-world applications. Led by Cambridge Institute of Technology Bangalore faculty and student researchers, the session covered:
- Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs)
- Transformer architectures and foundation models
- Transfer learning and fine-tuning strategies
- Ethical considerations in generative AI development
Professional Experience
Driving innovation in AI and machine learning through high-impact roles at industry leaders.
AI Specialist
Trilliant Digital Pvt Ltd- Designing and shipping production-grade LLM pipelines for AI-driven digital marketing campaigns, integrating Google Gemini 2.5, GPT-4o, and Claude via multi-provider routing with fallback logic.
- Building end-to-end N8N and MCP server automations that removed repetitive manual steps across SEO auditing, ad copy generation, and performance reporting workflows.
- Implementing RAG architectures with Supabase pgvector for knowledge-grounded AI outputs, reducing hallucination rates and improving content relevance.
- Leading research into multimodal AI applications for brand content generation using Stable Diffusion SDXL and IP-Adapter.
Agentic AI Intern
Trilliant Digital Pvt Ltd- Built multi-agent CrewAI pipelines with specialist agents for SEO analysis, ad copy generation, and competitor research - cutting manual effort by 60%.
- Designed LangChain RAG pipelines grounded in FAISS and Supabase vector stores, enabling accurate, source-cited AI marketing content at scale.
- Engineered custom API integrations and advanced prompt chains, achieving a 3x increase in content delivery speed.
- Built a Google Ads Campaign Copy Generator using Gemini 2.5 Pro and OpenRouter with multi-model fallback, producing structured ad groups with Excel export capability.
Team Lead
Samsung R&D Institute India, Bangalore (SRI-B)
Awarded for engineering a platform-agnostic Stable Diffusion pipeline that achieved 80% image quality accuracy while eliminating vendor lock-in across all internal generative AI workflows.
- Reverse-engineered AUTOMATIC1111 (A1111) WebUI dependencies to extract and run SDXL, ControlNet, and IP-Adapter modules as a standalone CLI pipeline without browser-based UI overhead.
- Implemented LoRA and DreamBooth fine-tuning workflows enabling rapid custom model adaptation on internal datasets with GPU-accelerated training.
- Achieved 80% image quality accuracy (SSIM/FID benchmarks) across diverse hardware environments, enabling cross-platform generative AI deployment.
- Built automated batch inference scripts reducing per-image generation time by 40% versus the baseline AUTOMATIC1111 setup.
Research Intern
Samsung R&D Institute India - Bangalore
- Built a multimodal Video Q&A chatbot using Video-LLaVA-7B for visual reasoning and OpenAI Whisper (large-v2) for speech-to-text, achieving 96% transcript accuracy.
- Ran GPU-accelerated hyperparameter tuning (learning rate, beam width, temperature) reducing inference latency by 35% while maintaining accuracy benchmarks.
- Integrated the chatbot with a Flask REST API, enabling real-time video query responses with streaming output support.
- This work established a new accuracy benchmark for multimedia AI at SRI-B and contributed to the subsequent promotion to Team Lead.
Technical Skills
A comprehensive toolkit spanning AI research, engineering, and deployment at scale.
AI & Machine Learning
Frameworks & Libraries
Tools & Platforms
Specializations & Languages
Projects
Production AI systems, research implementations, and open-source contributions across LLMs, multi-agent systems, and machine learning.
LLM Code Assistant
An AI-powered tool providing real-time, context-aware code suggestions for developers. Leverages LLM capabilities to understand code context and generate accurate completions and explanations.
Finance Agent
An intelligent AI agent for financial analysis, market research, and investment insights. Uses LLMs and tool-calling to retrieve, process, and reason over real-time financial data.
MECE Audience Segmentation
AI-powered audience segmentation applying MECE (Mutually Exclusive, Collectively Exhaustive) principles for precision marketing. Enables data-driven targeting without overlap or gaps.
EEG-Based Age & Gender Classification
A neuro-AI model classifying age and gender from EEG brainwave signals using deep learning. Combines signal processing, CNN architectures, and neuroinformatics for high-accuracy diagnostics.
ShadowFox - AI/ML
A collection of AI/ML projects exploring advanced algorithms, classification models, and intelligent prediction systems. Built during applied AI research with diverse real-world datasets.
Google Ads Campaign Copy Generator
Converts keyword lists into structured ad groups with professional copy using Gemini 2.5 Pro and OpenRouter's GPT-4o-mini as fallback. Outputs ready-to-import Excel files for Google Ads campaigns.
LangChain - Complete Guide & Projects
A comprehensive, hands-on guide to LangChain covering chains, agents, RAG pipelines, memory modules, and real-world application builds. Includes end-to-end project implementations with Python.
Amazon Lens
An AI-powered product intelligence tool that analyzes Amazon product listings, reviews, and market data. Extracts actionable insights for sellers using NLP and sentiment analysis pipelines.
Awards & Achievements
Milestones and recognition earned through innovative AI engineering and impactful research.
Outstanding Award
Received the prestigious Outstanding Award from Samsung R&D Institute India for engineering a platform-agnostic Stable Diffusion pipeline achieving 80% image quality accuracy.
Excellence Award
Recognized with the Excellence Award at Samsung R&D Bangalore for eliminating vendor lock-in and enabling GPU-accelerated generative AI workflows across diverse environments.
Springer Publication
Peer-reviewed research published in Springer Nature Singapore - SmartCom 2025 proceedings on Generative AI-Powered Video Summarization with CNN + LLM architectures.
96% Transcript Accuracy
Engineered an AI Video Q&A chatbot using Video-LLaVA-7B and Whisper achieving 96% transcript accuracy - setting a new benchmark in multimedia AI at Samsung R&D.
60% Effort Reduction
Automated core digital marketing workflows (SEO, Google Ads) using LangChain and CrewAI, delivering a 60% reduction in manual effort and 3× faster content delivery.
GenAI Workshop Speaker
Presented at the World Resources Webinar on Generative AI, representing Cambridge Institute of Technology Bangalore - covering GANs, Transformers, and ethical AI development.
Let's Connect
Open to opportunities, collaborations, and interesting conversations about AI. Let's build something remarkable.