I’m a Research Staff Member at IISc DREAM Lab and a Machine Learning Practitioner specializing in Cloud Computing and Scalable Agentic Systems. My work focuses on building infrastructure for Massive Agent Interaction Workflows, Distributed Agent Architectures, and Hybrid Cloud Applications (XFaaS).
Previously at Harman International (Samsung), I led initiatives in multi-agent frameworks and SecOps automation. I am deeply invested in bridging the gap between distributed systems and advanced AI, currently facilitating the training of large-scale language models on distributed clusters.
Research Staff – Cloud & Agentic Systems @ IISc DREAM Lab Jan 2026 – Present
-
Research Focus:
- Scalable Agents: Architecting systems for massive agent interaction workflows and distributed agent orchestration to enable high-concurrency autonomous operations.
- Memory & Infrastructure: Investigating Agentic Memory Systems at scale and developing Cross-Platform FaaS (XFaaS) for one-touch deployment of hybrid cloud applications (AWS/Azure).
-
Teaching Assistant (DS256: Scalable Systems for Data Science):
- Assisting Professor Yogesh Simmhan in guiding students through large-scale distributed computing concepts.
- Infrastructure & Curriculum: Orchestrating the lab’s inaugural Language Model pretraining pipeline. Guiding students in architecting data pipelines for 4TB of CommonCrawl data using Apache Spark on an HDFS cluster.
- Distributed Training: Supervising the implementation of PyTorch Distributed for training and inference of LLMs on cluster environments.
Associate Data Scientist – AI & ML Dept. @ Harman DTS Jan 2024 – Jan 2026
- Tech Lead (Gen-AI Invoice Processing): Architected a multi-agent system with Ray and Prefect, achieving optimized processing time via parallel execution. Fine-tuned Qwen 2.5 VL (3B) using QLora to achieve 78% F1-score.
- Tech Lead (SecOps Copilot): Developed an incident triaging agent integrating Microsoft Sentinel & MISP, reducing Mean Time to Respond (MTTR) by 40%.
- Mentorship: Conducted AI agent development sessions for 100+ engineers.
| Domain | Technologies |
|---|---|
| Languages | Python, C++, C, Java, SQL |
| Distributed Systems | Apache Spark, HDFS, PyTorch Distributed, Ray, Prefect |
| GenAI & Agents | LangChain, LangGraph, Autogen, LlamaIndex, Ollama, Hugging Face |
| Cloud & MLOps | XFaaS, AWS, GCP, Azure, Docker, Kubernetes, MLFlow |
| Databases | Qdrant (Vector), Neo4j (Graph), PostgreSQL, MongoDB |
| Backend | FastAPI, Django, RESTful Services, SSE Streaming |
-
[PokeAgent - NeurIPS/Google DeepMind Challenge]
-
Leading a team to develop a VLM-powered autonomous agent for long-horizon planning.
-
Tech: Neo4j (Lore Mapping), NetworkX (Task Planning), Custom A* Planner.
-
Focus: Generalized HTN (Hierarchical Task Network) planning frameworks.
-
[English-to-Hinglish Translation]
-
Trained Seq2Seq Transformer models on code-mixed datasets, reducing training time from 30 mins to 2 mins while eliminating repetition.
-
[IEEE Publication: Deep Learning in Fashion]
-
Proposed a CNN-GAN architecture for virtual try-on and neural body fit estimation.
- Be Brilliant Rising Star Innovation Award (Sep 2025) – Recognized by Harman DTS for R&D in multi-agent systems.
- Co-Founder & CTO @ Olivia Family Restaurant – Built an AI-powered billing system reducing validation time by 96%.
- Solving for India Hackathon Institutional Winner – Geeks for Geeks / Google / AMD.
- AWS AI Practitioner
- Generative AI with LLMs – DeepLearning.AI
- Microsoft Azure AI-900
Feel free to connect with me on LinkedIn