BTech in Artificial Intelligence & Data Science · 2 years as an AI Engineer. I build production-ready systems: multi-agent LLM pipelines, generative AI (LoRA training, diffusion, video generation), and full-stack AI products. Based in India.
🤖 Agentic & LLM Systems — Multi-agent pipelines (LangGraph), RAG with structured evidence passing, fine-tuning with SFT + GRPO on Qwen and Llama models
🎨 Generative AI & Vision — LoRA training pipelines (Kohya_ss → SD / FLUX.1), text-to-video (Wan2.2), computer vision (CLIP · RetinaFace · YOLOv5)
🏗️ Full-Stack AI Products — FastAPI + Next.js + PostgreSQL, deployed apps on Render, deterministic safety layers over LLM outputs
📊 Data & Financial ML — Time series forecasting, crypto/retail analytics on Snowflake, clustering and EDA pipelines
agent orchestration planner-executor patterns multi-agent debate tool/function calling RAG evaluation query rewriting hybrid retrieval embedding search evidence attribution structured outputs guardrails prompt routing memory design human-in-the-loop review LLM observability self-scoring agents
parameter-efficient fine-tuning LoRA/adapter training reward modeling preference optimization diffusion pipelines text-to-image workflows image-to-video workflows identity preservation virtual try-on face analysis vision-language classification dataset curation inference optimization GPU/ROCm workflows
| Project | What it does | AI skills |
|---|---|---|
| Multi-Agent RAG | Three-agent literary consistency pipeline: claim decomposition → evidence retrieval → contradiction judging. Structured Python object passing preserves similarity scores and claim provenance across agents. 98.3% success rate on 60 test cases. IIT Kharagpur Hackathon | Multi-agent orchestration, atomic fact decomposition, semantic retrieval, contradiction detection, evidence provenance, RAG evaluation |
| BTC-Forecaster | Fork of TradingAgents repurposed for intraday BTC forecasting. Multi-agent pipeline: technical + news + sentiment analysts → bull/bear debate → trader → risk → final forecast. Emits 1h/4h predictions with a self-scoring track record logged against realized prices. | Agentic market research, debate-based reasoning, time-series signal fusion, sentiment aggregation, risk-aware decisioning, forecast backtesting |
| AMD AIPL | Fine-tuned an LLM with SFT + GRPO for a 1v1 Q-agent vs A-agent tournament. Custom reward function; self-play loop where question and answer models iteratively improve each other. AMD Hackathon at IIT Bombay | Supervised fine-tuning, GRPO, reward shaping, self-play, agent evaluation, adversarial QA |
| Project | What it does | AI skills |
|---|---|---|
| AI-Avtaar | End-to-end character pipeline: upload photos → automated LoRA training → image generation → virtual clothing try-on. Four isolated Python environments orchestrated through a single Streamlit UI. | Identity-preserving generation, LoRA training, dataset preprocessing, diffusion workflow orchestration, virtual try-on, UX for model pipelines |
| AI Video Creator | Three-stage generation pipeline: storyboard planning → per-scene image generation → text-to-video animation. Gradio tabbed interface. Tested on AMD MI300X with ROCm. | Story-to-scene planning, prompt engineering, diffusion chaining, image-to-video orchestration, GPU inference, creative AI tooling |
| Gender Detection API | FastAPI endpoint for face detection and vision-language gender classification. Handles multiple faces, non-human images, and mismatches as distinct typed error responses. | Computer vision inference, face detection, zero-shot image classification, typed error design, API deployment |
| Project | What it does | AI skills |
|---|---|---|
| ResumeTeX | Browser-based LaTeX resume builder: manual form, AI import from PDF/DOCX (extract → parse → verify pipeline), AI tailoring to job descriptions. Deterministic anti-fabrication guard restores all original facts post-AI. Deployed on Render. | Document AI, information extraction, structured parsing, factuality guardrails, AI-assisted rewriting, production full-stack AI |
| Stock News Summarizer | Scrapes market-news sources; an LLM selects top articles and writes sub-500-word summaries with 7-day "what changed today" diffs. SQLite history, daily refresh at 8 AM IST. Free-tier deployed on Render. | News ranking, abstractive summarization, temporal diffing, scheduled AI workflows, retrieval over history |
| Student Performance Analysis | CSV upload → K-Means / Agglomerative clustering → per-student performance dashboard with trend charts, subject breakdowns, class comparisons, and Excel export. | Unsupervised learning, educational analytics, feature preprocessing, cluster interpretation, ML dashboards |
| Project | What it does | AI skills |
|---|---|---|
| interactive-preview-skill | Agent skill that turns React/Next.js codebases into interactive, theme-matched "try it before you sign up" demos with guided product tours on mock data. Leaks no backend. | Agent instruction design, codebase analysis, UI generation, safe mock-data workflows, developer tooling |
| Support Finder | MV3 Chrome extension with a four-layer deterministic pipeline: DOM scan → same-domain path probing → schema.org extraction → confidence scoring. Returns ranked support contacts with explanations. No AI; no fabrication. | Heuristic extraction, confidence scoring, explainable ranking, deterministic information retrieval, browser automation |
| android-compose-design | Agent skill for mobile UI generation. Guides AI to produce distinctive Jetpack Compose UI with intentional color, type hierarchy, shape language, and motion instead of Material 3 defaults. | Agent prompt architecture, design-system reasoning, mobile UI generation, style critique, creative coding guidance |
| Project | What it does | AI skills |
|---|---|---|
| Solana Price Analysis | OHLCV data (2021–2024), 44-column technical-indicator feature set, ML price prediction model, and live Binance price dashboard. | Financial feature engineering, technical indicators, price prediction, model evaluation, live analytics |
| Rossmann Retail Analysis | 1M+ row sales dataset: cleaning, feature engineering, EDA on Snowflake. Quantified 81.5% sales uplift from promotional periods. | Large-scale EDA, feature engineering, SQL analytics, retail forecasting signals, business insight extraction |
| Time Series Forecasting | Four-framework side-by-side: LSTM on temperature data, airline passengers, stock SARIMAX, and stock LSTM. Flask web interface for the SARIMAX model. | Sequence modeling, SARIMAX forecasting, comparative model evaluation, regression metrics, ML web serving |
