|
|
|
|
|
||
AI/ML Engineer (final-year B.Tech, AI & Data Engineering @ LPU), currently an AI Engineering Intern working on LLM Systems & Applied GenAI — building production-grade agentic pipelines with LangGraph, open-weight LLMs, and full observability.
- Building EquityFlow, a multi-agent equity research system (LangGraph, Llama-3.1-8B, Tavily, yfinance, ChromaDB, FastAPI, LangFuse).
- Author of AutonoML v2.3.0, a multi-agent AI platform — Planner (DAG), parallel Research, ReAct Execution, hybrid LLM + programmatic Evaluation, two-tier memory, 386 tests; SmartRouter cuts latency from ~45s → ~2s on direct queries.
- Previously built data/AI pipelines at Futurense Technologies.
- Shipped AIOps RCA (XGBoost + SHAP + FastAPI + Docker) and a serverless Model Arbitration Engine on AWS (87ms avg latency across sklearn / XGBoost / Bedrock).
- Interests: LLM orchestration & agents, RAG evaluation, MLOps/AIOps, GPU-poor-friendly open-source model engineering.
| Languages |
|
| AI / ML & GenAI |
|
| Data & Backends |
|
| Cloud & DevOps |
|
   
 
|
   
|
| Project | Description | Stack |
|---|---|---|
| EquityFlow · Multi-Agent Equity Research | Multi-agent equity research system — research, writer, and critic agents over live web + financial data with persistent memory and full tracing. | LangGraph Llama-3.1-8B ChromaDB FastAPI LangFuse |
| AutonoML · Multi-Agent AI Platform | Planner (DAG) + parallel Research + ReAct Execution + hybrid LLM/programmatic Evaluation; two-tier memory, self-reflection loop, 386 tests; SmartRouter: ~45s → ~2s on direct queries. | FastAPI Streamlit ChromaDB Ollama/Groq |
| LoopForge · Agentic Loop Platform | LangGraph multi-loop agentic platform with distributed task execution and production observability. | LangGraph Celery Redis PostgreSQL LangFuse |
| AIOps RCA System · Root Cause Analysis | XGBoost root-cause classifier over LEMMA-RCA incidents — 96 features, 6 classes, SHAP explainability, cascade-risk detection, Go/No-Go deployment gate. | XGBoost SHAP FastAPI PostgreSQL Docker |
| Model Arbitration Engine · AWS Serverless | Real-time router across sklearn / XGBoost / Bedrock Haiku — epsilon-greedy + EMA latency tracking; 87ms avg latency at 95.3% accuracy. | AWS Lambda SageMaker Bedrock DynamoDB |
| Ollive · LLM Assistant Benchmark | Llama-3.1-8B vs Llama-3.3-70B benchmark with LLM-as-judge scoring, 4-layer guardrails, and LangFuse v4 tracing; deployed on Render + HF Spaces. | Groq LangFuse Streamlit Render |
AI/ML Engineer in the making · B.Tech AI & Data Engineering @ LPU · ex-Futurense · Agentic AI, LLM Systems & MLOps