Agam Pandey, Applied AI Researcher and Member of Technical Staff at Mem0

Agam Pandey

Applied AI Researcher · Member of Technical Staff at Mem0

I build and study memory, retrieval, and evaluation systems for AI agents. I’m an Applied AI Researcher and Member of Technical Staff at Mem0, and a 2026 graduate of IIT Roorkee. My work spans agentic memory, retrieval-augmented generation, semantic search and reranking, LLM evaluation, and multimodal safety for production AI systems.

Previously, I was a Founding Engineer at Scriben, backed by South Park Commons, and worked at Swiggy, Akki AI, EzAIx, Mindshift Analytics, and Strabag Efkon India. My work has appeared at ICML-SCALE, NeurIPS-RML, MLRC, TMLR, and AAAI-SAPP. Off the clock, I run middle/long distance with Athletics Team IIT Roorkee and contribute to Data Science Group.

Role
Applied AI Researcher · MTS
Org
Mem0 · ex-Scriben · ex-Swiggy
Focus
Agentic memory · information retrieval · evals · multimodal safety
Venues
ICML · NeurIPS · AAAI · TMLR
Education
IIT Roorkee · Class of 2026

News

Scroll
  • Joined Mem0 as an Applied AI Researcher and Member of Technical Staff.
  • Concluded my tenure as Founding Engineer at Scriben.
  • Scriben showcased at the South Park Commons Summer 2026 Demo Faire, San Francisco. Recap.
  • Joined Scriben as Founding Engineer, backed by South Park Commons.
  • Presented "Before It Persists" as a poster at the SCALE Workshop, ICML 2026 (Seoul). LinkedIn post.
  • Paper "Before It Persists: Write-Time Defense for Multimodal Agent Memory" accepted at SCALE Workshop, ICML 2026.
  • Started as Reviewer for SCALE Workshop ICML 2026.
  • Received the Urmila & Prem Nath Kumra Annual Excellence Award from the IIT Roorkee Heritage Foundation USA. View award.
  • Started as a Reviewer for an ICLR Workshop.
  • Joined mem0 as an Applied AI Research Intern.
  • Released DL-Playground v1, an open-source visual PyTorch graph builder, with DSG, IIT Roorkee and SDSLabs. Live demo.
  • Reviewer as Program Committee Member for AAAI-26 Student Abstract and Poster Program (SAPP).
  • Presented ReCroPA at MLRC, Princeton University (New Jersey, USA).
  • Started as AI Engineering Intern at Swiggy (CEO's Office), Bangalore.
  • Promoted to Joint Secretary at Data Science Group, IIT Roorkee.
  • Promoted to Vice President (Analytics) at Society of Business, IIT Roorkee.
  • Joined EzAIx Inc. as AI Engineer Intern, working on multi-agent systems.
  • Started as AI Engineer Intern at Akki.ai, building multi-agent copilot systems.
  • Started as Research Intern at Transportation Engineering Department, IIT Roorkee under Prof. Pushpa Choudhary.
  • Joined Mindshift Analytics as Data Science Intern.
  • Appointed as Joint Secretary of Athletics, IIT Roorkee, managing 60+ athletes.
  • Started internship at EFKON India as Machine Learning Engineer.
  • Promoted to Manager (Analytics) at Society of Business, IIT Roorkee.
  • Became Member at Data Science Group.
  • Joined Hanomi as ML Intern, contributing to AI-powered hardware design acceleration.
  • Started as MLOps Engineer at Tinkering Lab, IIT Roorkee, working on AlgoEase startup.
  • Appointed as Village Documentation Coordinator at Unnat Bharat Abhiyan, IIT Roorkee.

Service & Activities

Selected Projects

Research

  • StegGNN

    Graph neural network operating on patch graphs for learned image steganography.

  • KGE-CLIP

    Knowledge-graph embeddings as a retrieval signal for GraphRAG; CLIP-style text-graph alignment.

  • Hybrid IR Recommender

    BM25 and dense (BGE) retrieval fused via weighted RRF, with an LLM query planner and cross-encoder reranking.

  • Vehicle Motion Forecasting

    Attention over spatial neighbors and temporal history for multi-agent vehicle motion forecasting.

Systems & Applied

  • DL-Playground

    Open-source visual PyTorch graph builder with live shape/FLOP feedback and code emission.

  • ByeLabs

    Local-first multi-agent pipeline for healthcare roster extraction and normalization; runs fully on-device.

  • Zomato Nugget

    Multimodal agentic GraphRAG over restaurant menus and reviews for grounded recommendation.

  • AI Interview Platform — Hackathon

    QLoRA-fine-tuned Mistral-7B for adaptive interview questions, with Whisper ASR and real-time speech feedback.

Research & Publications

ReCroPA paper figure

Revisiting CroPA: Enhancing Cross-Prompt Adversarial Transferability Best Paper Award

A. Pandey et al. · TMLR / MLRC 2025
Paper MLRC TMLR
Before It Persists paper figure

Before It Persists: Write-Time Defense for Multimodal Agent Memory

A. Pandey · ICML 2026 SCALE Workshop
Paper Project Page Workshop
CroPA++ paper figure

CroPA++: Exposing Vulnerabilities in Vision-Language Models

A. Pandey et al. · NeurIPS Reliable ML Workshop 2025
Paper Workshop
Scene Graph CoT figure

Zero-Shot Vision Language Reasoning via Dual-layer Scene Graph CoT

AAAI-26 Student Abstract
Paper Conference
TransPatch universal adversarial patch figure

TransPatch: Learning a Universal Adversarial Patch for ViT-CNN Cross-Architecture Transfer in Semantic Segmentation

AAAI-26 Student Abstract
Conference
StegGNN paper figure

StegGNN: Learning Graphical Representation for Image Steganography

ICCV CV4DC Workshop (under revision)
Manuscript Code

Experience

Scroll
  • Applied AI Researcher (Member of Technical Staff) | Mem0
    Oct 2026–Present

    Working on agentic memory, information retrieval & search, and evals & benchmarking for production AI systems.

  • Founding Engineer | Scriben
    San Francisco · Jul–Sep 2026 · backed by South Park Commons

    Built Scriben’s algorithms, systems, application stack, and production infrastructure; owned engineering and product. Supported investor demos, fundraising, GTM, and early revenue.

  • Applied AI Research Intern | Mem0
    Remote (San Francisco) · Jan–Jul 2026

    Worked on agentic memory algorithms, information retrieval & search, evals & benchmarking, and continual learning. Used DPO fine-tuning and knowledge distillation for extraction and retrieval models.

  • AI Intern | Swiggy (CEO's Office)
    Bangalore · May–Aug 2025

    Built multi-agent conversational recommender for Swiggy Scenes using LangGraph and multimodal GraphRAG.

  • AI Engineer Intern | Akki AI
    Remote (New Delhi) · Dec 2024–Apr 2025

    Built a startup copilot with multi-agent CUA workflows (LangGraph) and Hybrid-RAG with persona memory. RLHF-style evaluation and deployed scalable agent clusters on AWS (FastAPI + Celery).

  • AI Engineer Intern | EzAIx Inc.
    Remote (Florida US) · Mar–Apr 2025

    Multi-agent onboarding assistant for Microsoft Suite workflows: GPT-4V + LangGraph with RAG, async orchestration (FastAPI/Celery, MongoDB), and Selenium UI control.

  • Research Intern | Transportation Lab, IIT Jodhpur
    Aug 2024–Present · with Prof. Akshay Gupta

    Depth estimation and multimodal sensor fusion (16-beam LiDAR + monocular camera) for early collision-warning systems in Indian traffic.

  • Data Science Intern | Mindshift Analytics
    Remote · Aug–Oct 2024

    Built telemetry analytics for mining fleets: fuel/dynamics modeling and anomaly detection at scale. Implemented RDP trajectory simplification + DBSCAN clustering pipelines.

  • Applied ML Intern | Strabag Efkon India
    Gurgaon · Jun–Jul 2024

    ML + linear-programming (PuLP) fleet optimization for emergency-response routing on Hyderabad-ORR; modular Python tooling later scaled to other highways.

Athletics

800 m
2:12

Personal best

1500 m
4:36

Personal best

5000 m
17:02

Personal best

Agam Pandey racing the 800 m
Agam Pandey racing the 1500 m
Agam Pandey racing the 5000 m