Open to international research & engineering roles

Architecting intelligent systems from theory to deployment.

AI Architect · Generative AI · Computer Vision · Robotics

10+ years turning cutting-edge research — VLMs, RAG, diffusion models, and Vision-Language-Action systems — into production-grade products across manufacturing, energy, finance & document intelligence. Founder of EulerDB.

Start a Collaboration Download CV
Prashant Verma
VLA Robotics
Advanced R&D
Founder, EulerDB
Graph AI infra
10+
Years across research & production AI
3
Peer-reviewed & arXiv publications
14
Business units served by GenAI platform
IIT
MTech, Data Science & AI · Patna
Profile

Bridging frontier research and
real-world engineering.

I'm an AI Architect and technical lead with a decade of multidisciplinary experience spanning generative AI, computer vision, robotics, and graph theory. My work runs the full arc — from unsolved academic problems to enterprise-scale systems deployed in manufacturing, energy, finance, and document intelligence.

I've architected multilingual GenAI platforms fusing LLMs and VLMs, built agentic RAG pipelines that cut latency and cloud cost by double digits, and now lead Vision-Language-Action (VLA) research translating perception and language into precise robotic control.

I care deeply about graph-theory problems and cognitive architectures, and I'm actively seeking senior research or engineering roles — and international collaborations — where creativity and rigor drive strategic impact.

Noida, India
Open to relocation & remote
Technical Specialist, Advanced Robotics
HCLTech
Global delivery experience
India · Saudi Arabia · enterprise clients
3 publications · arXiv + IJARCCE
Document AI & graph theory
Core Competencies

What I build

A vertically integrated skill set — from model research and fine-tuning to edge deployment and cross-functional leadership.

Generative AI & LLMs

Agentic systems, RAG at scale, LLM/VLM fine-tuning, diffusion models, and Model Context Protocol integrations.

RAGAgenticLLMVLMDiffusionMCP

Computer Vision

Video analytics, OCR & document processing, object detection, 3D vision, face recognition, and image restoration.

YOLOOCR3D VisionSigLIP-2BLIP-2

Robotics & VLA

Vision-Language-Action models, perception-driven control, and robotic manipulation pipelines for dynamic environments.

VLAManipulationPerceptionGrasping

Graph & Data Systems

High-performance graph databases (EulerDB), GraphML, Neo4j, and novel heuristics for classic graph-theory problems.

EulerDBNeo4jGraphMLVertex Coloring

Cloud & MLOps

OCI & SAP OCI, ETL orchestration with Prefect, edge inference on Jetson, and Dockerized production pipelines.

OCIPrefectDockerJetsonETL

Technical Leadership

Leading cross-functional teams, stakeholder alignment, research direction, and mentoring across global engagements.

Team LeadStrategyMentoringR&D
Trajectory

Experience

Selected roles delivering measurable impact across robotics, energy, and enterprise AI.

Technical Specialist — Advanced RoboticsMar 2026 — Present
HCLTech · India
  • Leading R&D of Vision-Language-Action (VLA) models for robotic manipulation.
  • Building pipelines that translate visual and language inputs into precise control actions, improving grasping and task execution in dynamic environments.
  • Translating state-of-the-art vision-action research into production-grade robotics capabilities.
Lead Data Scientist — Data Science & AIAug 2025 — Feb 2026
EPAM Systems & ARAMCO · Saudi Arabia
  • Built a heat-transfer fouling-factor microservice enabling real-time monitoring and predictive maintenance for refinery heat exchangers.
  • Architected an ETL pipeline for Material Resource Optimization on the Supply Chain Digital Twin using Prefect orchestration and SAP OCI extraction.
Lead Data Scientist / Technical Lead — AI, CV & AnalyticsMay 2022 — Aug 2025
L&T Technology Services & SABIC · Saudi Arabia
  • Designed a company-wide GenAI architecture fusing LLMs (GPT-4-Turbo, Llama 3) with VLMs (SigLIP-2, BLIP-2) for multilingual retrieval across 14 business units.
  • Built an agentic, priority-aware RAG pipeline — reduced latency 37% and cloud cost 42%.
  • Boosted industrial-document OCR accuracy from 68% to 95%; recovered degraded engineering drawings via fine-tuned Stable Diffusion inpainting.
  • Architected an Oil & Gas asset-health CV system with YOLOv8 defect detection at 92% mAP, cutting manual inspection 60%.
Computer Vision Engineer — CV, 3D Vision & Real-Time AnalyticsApr 2019 — Apr 2022
HCL Technologies · Noida, India
  • Engineered a proctoring-malpractice detection system for PearsonVUE using RetinaNet, with real-time WebRTC streaming for 10,000+ concurrent users.
  • Trained a conditional GAN for LiDAR up-sampling (4× density, ±2mm) and an ArcFace face-ID engine reaching 98.6% accuracy on LFW.
  • Built GPU-accelerated live video stitching (320ms → 110ms) with ResNet-18 human re-identification.

Earlier: DAAS Labs · Zvesta · National Institute of Medical Statistics (HIV analytics with WHO/UNAIDS) · IASRI–ICAR · GreySugar (Founder) · IISER Kolkata — full history in CV.

Publications

Research

Peer-reviewed and preprint work spanning document intelligence and graph theory.

01

S2 Chunking: A Hybrid Framework for Document Segmentation Through Integrated Spatial and Semantic Analysis

arXiv:2501.054852025Document AIFeatured on Hugging Face
Read paper
02

Degree-Based Logical Adjacency Checking (DBLAC): A Novel Heuristic for Vertex Coloring

arXiv:2501.124792025Graph TheoryBenchmarked vs DSATUR / RLF
Read paper
03

High Dimensional Object Analysis Using Rough-Set Theory and Grey Relational Clustering

IJARCCE2016Pattern Analysis
Find
Open Source

Code in the wild

Selected repositories on github.com/Vprashant — from the S2-Chunking reference library to generative-AI frameworks.

s2-chunking-lib 15

Structural-semantic chunking of documents — the reference implementation of the S2 Chunking paper.

Python
Eulerchain 1

A generative-AI framework — part of the EulerDB graph-AI ecosystem.

Python
LangGPTEval 1

Evaluation library assessing LLM response quality on faithfulness and relevancy.

Python
fancybbox 1

Fancy bounding-box rendering utilities for object-detection visualization.

Python
Gstreamer_FaceDetection 1

Real-time face detection built on the GStreamer media pipeline.

Python
global_logistic_rag 0

Retrieval-augmented generation system for global logistics knowledge.

Python
Foundations

Education & recognition

Education

2024 — 2026
MTech, Data Science & Artificial Intelligence
Indian Institute of Technology (IIT), Patna
2013 — 2015
MTech, Computer Science & Engineering
RGP University, Bhopal
2007 — 2011
BTech, Computer Science & Engineering
UIET, CSJM University, Kanpur

Awards & Certifications

Awards

Spot Award — L&T (2023) ERS Champion — HCL (2021) Winner — HCL Hackathon (2019)

Certifications

Google Cloud — Generative AI & Transformers (2023) Udemy — Python & R for ML/DL (2017)
Let's build together

Open to international collaboration.

Research partnerships, senior engineering roles, or ambitious AI & robotics problems — I'd love to hear from you.

Prashant27050@gmail.com Connect on LinkedIn