AI Engineer building enterprise-grade agentic systems, multi-modal pipelines and intelligent automation. Turning cutting-edge research into production-ready solutions.
Prodapt R&D
Sunrise GmbH
Tech Mahindra
University of Edinburgh
VIT Vellore
Prodapt R&D
Sunrise GmbH
Tech Mahindra
University of Edinburgh
VIT VelloreI'm an AI Engineer currently pursuing my Master's in Artificial Intelligence at The University of Edinburgh. My work lives at the intersection of cutting-edge research and real-world enterprise AI.
At Prodapt's NextGen Labs (R&D), I was part of a team of 50+ developers focused on researching and integrating cutting-edge GenAI tools into enterprise workflows. I personally built 30+ modular AI agents covering everything from voice navigation to medical data digitisation.
Agentic AI, LLM orchestration, RAG, Vision-Language Models, NL2SQL — applied to real-world problems.
English (Professional · IELTS C1) · Hindi (Native) · Japanese (Basic)
From dissertation research to hackathon finalists — building AI that matters.
Agentic chatbot that translates natural language queries into SQL to retrieve live data from the IUPHAR Guide to Pharmacology PostgreSQL database — serving academic researchers and drug discovery teams. Evaluating LangGraph vs CrewAI pipelines with a PostgreSQL MCP server for schema-aware access.
AI-powered Chrome extension with three real-time tools: voice-controlled agentic navigator, social cue coaching assistant, and autonomous documentation engine. Multimodal pipeline combining Whisper + GPT-4o Vision for live lecture capture. Low-latency architecture using WebSockets and GPT-4o function calling.
Fine-tuned & benchmarked CLIP, BLIP, BLIP-2, OFA on FashionGen dataset using LoRA, full fine-tuning, and mapping-network tuning. Accelerated training 250%+ via multi-GPU parallelisation. Unified evaluation across CIDEr, BLEU-4, METEOR, ROUGE-L.
LangGraph-orchestrated agent interpreting natural language instructions and executing them via MoviePy. Trim, crop, mute/bleep, blur, generate subtitles, and translate — all through chat. Deployed as a production API endpoint for enterprise use.
Trained YOLOv8 from scratch on a combined Kaggle + Roboflow dataset spanning 59 unique traffic sign classes. Deployed as a real-time Streamlit app with live camera input for detection.
LangGraph-orchestrated multi-agent system: CrewAI for billing, AutoGen for troubleshooting, LangChain for recommendations. LlamaIndex managed hybrid SQL + vector retrieval from telecom docs and SQLite DB.
Selected among top finalists from 3,000+ participants across Prodapt in a multi-stage AI innovation competition. Competed in the final round against the best in the company.
Placed Top 8 at the University of Edinburgh's AccessAI Hackathon, building a real-world AI accessibility solution for disabled users — competing against 35+ teams from across the university.
Personally designed and deployed 30+ modular AI plugin agents as production API endpoints across diverse enterprise client environments at Prodapt R&D — spanning voice, vision, data, and automation.
I'm open to AI research collaborations, internships, and interesting projects. Connect on LinkedIn or send me a message — no spam, just good conversations.