AI/ML Engineer + Generative AI Developer

AI/ML Engineer Building LLM Agents and Applied AI Systems

I am a 3rd-year B.Tech CSE student at Birla Institute of Applied Sciences, Bhimtal, focused on generative AI engineering, agentic AI, retrieval-augmented generation (RAG), natural language processing (NLP), computer vision, satellite imagery analysis, and scalable machine learning applications. I apply Python, PyTorch, LangChain, transformer fine-tuning, and MLOps best practices to turn research papers and real-world problem statements into recruiter-ready AI products with clean interfaces, measurable business impact, and cloud deployment-ready engineering.

View projects Get in touch

Focus: GenAI, RAG, LLM Agents
Research: 10-Agent LLM Systems
Stack: Python, LangChain, FAISS

About

AI engineer with strong research execution and product delivery skills.

I work at the intersection of applied machine learning, generative AI engineering, and research-driven product development — turning complex model capabilities into software that actually ships and scales.

My focus areas include large language model (LLM) systems, retrieval-augmented generation (RAG), multi-agent architectures, and end-to-end deep learning pipelines. I'm proficient in Python, PyTorch, TensorFlow, and the modern NLP stack, with hands-on experience across the full model lifecycle — from experimentation and fine-tuning to evaluation, deployment, and monitoring.

Recent projects span private document QA systems, memory-augmented conversational AI, agentic workflow automation, physics-guided remote sensing models for satellite imagery, and CLIP-based recommendation engines for product intelligence. I regularly apply transformer fine-tuning, LoRA/QLoRA, prompt engineering, vector search, and tool-use patterns to build production-ready AI systems on AWS and Docker.

I care as much about clean software design and user experience as I do about model performance — because a great model that nobody can use isn't a solution.

Experience

Applied AI Engineering Experience

AI Research and Development Trainee

DIBER & DIPAS, DRDO

Developing ML-driven predictive models to optimize energy allocation for hybrid power and energy management systems, supporting defense-grade reliability and efficiency requirements.

Building predictive models for real-time energy allocation and load forecasting in hybrid power systems.
Designing optimization pipelines to balance energy sources and improve power management efficiency.
Translating research findings into validated, demo-ready engineering prototypes for DRDO applications.

AI Agent Development Intern

Shell Global Pvt. Ltd. | Remote | Dec 2025 - Feb 2026

Built production-grade multi-agent AI systems with LangChain and CrewAI, improving workflow automation and reducing manual processing by 65%.

Engineered autonomous tool-calling agents using GPT-4, Llama, prompt engineering, and task orchestration.
Reached 88% task accuracy across agentic workflow evaluations and reliability tests.

Generative AI Intern

Analytics Vidhya | Remote | Nov 2025 - Dec 2025

Developed retrieval-augmented generation pipelines and scalable AI APIs using FAISS embeddings, document retrieval, and LLM integration patterns.

Processed 10K+ documents with 92% retrieval accuracy across semantic search workflows.
Built scalable APIs serving 500+ concurrent users for AI-powered document intelligence.

Projects

AI, ML, and LLM Engineering Projects

Satellite AI

Generative AI-Based Cloud Removal for LISS-IV Satellite Imagery

Physics-guided computer vision and deep learning system using SAR-optical fusion and NDVI preservation, evaluated through PSNR and SSIM.

Impact focus: preserves vegetation signals while reconstructing cloud-obscured satellite regions.

PythonPyTorchRemote SensingSAR Fusion

View on GitHub

LLM Systems Research

Hybrid Multi-Agent Conversational AI Architecture

First-author LLM research project introducing a 10-agent modular AI framework with supervisor routing, competitive debate, and expertise-aware personalization.

Impact: +11.9% accuracy over a single-model baseline and 31.4% fewer unnecessary tool calls.

PythonCrewAILlama 3.3LLM-as-a-Judge

View on GitHub

Memory Systems

Memory-Augmented LLM using LSTM

Hybrid AI architecture combining Transformer reasoning with LSTM-based temporal memory for long-term conversational context and memory-aware assistant behavior.

Key idea: separate reasoning from sequential memory instead of expanding context windows.

PyTorchLSTMTransformersEmbeddings

View on GitHub

Agentic Systems

CivicAI

Streamlit civic complaint management platform with Leaflet/Folium maps, OSRM GPS routing, and a multi-agent AI backend deployed on Render.

Impact focus: coordinates geospatial routing with AI triage, automation, and workflow prioritization.

StreamlitFoliumOSRMAgents

View on GitHub

Security

Adaptive AI Steganography

Streamlit cybersecurity app for texture-guided LSB embedding with Fernet and PBKDF2 encryption for controlled hidden-message workflows.

Key challenge: balancing payload capacity, visual quality, and encryption hygiene.

PythonStreamlitFernetPBKDF2

View on GitHub

Product Intelligence

Fashion Product Intelligence System

CLIP-based AI recommendation engine with FAISS-powered vector similarity search and deduplication for fashion product discovery.

Impact focus: turns visual embeddings into practical catalog search and product intelligence.

CLIPFAISSPythonEmbeddings

View on GitHub

Private RAG

Private RAG Document QA System

Secure enterprise-style document QA system with hybrid dense and sparse retrieval, Multi-Query RAG, sentence-transformers, Gemma-2B, and a Streamlit interface.

Impact: processed 5K+ queries, reached 90% relevance, and reduced latency from 3s to 0.4s.

LangChainFAISSGemma-2BStreamlit

View on GitHub

Agentic Research

Agentic AI Multi-Agent System for EV Research

Supervisor and co-worker agent system with 5+ specialized LLM agents for EV market intelligence, policy analysis, business research, and automated reporting on AWS.

Impact: reduced manual research effort by 70% through delegated agent workflows.

PythonAgnoGroqAWS

View on GitHub

Open Source Contributions

GitHub Contribution Activity

A consistent record of engineering activity across AI/ML projects, research prototypes, and open-source tooling.

Consistent Commits

AI / ML Repositories

Active Since 2023

Research + Products

Contribution Graph — github.com/Adirohansuyal

… contributions Last 12 months

Loading contributions…

Longest Streak — Loading…

Recent Commits

Fetching latest commits…

Skills

Tools, Frameworks & Technologies

Machine Learning & Deep Learning

Python PyTorch TensorFlow scikit-learn OpenCV CLIP CNNs RNNs LSTMs Transformers Fine-tuning LoRA QLoRA Model Evaluation NLP Computer Vision

Generative AI & LLM Engineering

LangChain LlamaIndex CrewAI Autogen LangGraph RAG Pipelines Embeddings Vector Databases Prompt Engineering Agentic AI Multi-Agent Systems LLM Orchestration

LLM Stack & Models

GPT-4 Gemini Pro Llama 2/3 Mistral Stable Diffusion Instruction Tuning RLHF Tool Calling Agent Orchestration LLM Evaluation

Data, Deployment & MLOps

FAISS Pinecone ChromaDB MongoDB SQL AWS EC2 AWS Lambda AWS S3 Docker Kubernetes REST APIs MLOps CI/CD Model Serving Cloud Deployment

Available for Opportunities

Get In Touch

Let's build something meaningful together.

Open to AI/ML engineering internships, generative AI roles, NLP and deep learning positions, research collaborations, and applied AI projects.

AI/ML Engineering Generative AI NLP & Deep Learning Research Collaborations

Email Me LinkedIn GitHub