VATSALYA BHADAURYA
Humanoid Robotics · AI Engineer · Researcher
0%
AI Engineer @ iHub Research & Robotics

नमस्ते, I'm Vatsalya Bhadaurya

I build real-world humanoid robotics systems: perception, behavior, and embodied AI pipelines deployed on physical robots using ROS2, Jetson, and VLM / VLA reasoning. My focus is robots that work in the real world, not simulations.

Download Résumé View Work
0Publications
0Product-Ready Projects
0Awards & Honors
0Years Building

01 About

Vatsalya Bhadaurya with a humanoid robot

I'm an AI engineer specializing in multimodal perception, on-device inference, and robotic intelligence. My journey in technology began in grade 8 and grew into a focused practice in applied robotics and embodied AI, building systems deployed on real robots, not simulations.

I develop AI stacks for humanoid robots on Jetson platforms using VLM / VLA pipelines, embedding models into real-world behaviors for healthcare and industrial domains. I've built industrial vision systems for automotive manufacturing where robustness and reliability are non-negotiable.

I'm an IEEE-published researcher focused on multimodal sensing and real-time ML. Alongside this, I founded Nextgen Research Lab & Infrastructure Development Pvt. Ltd., where we're building MaaKosh, an AI-powered maternal & child healthcare ecosystem for preventive, accessible, and intelligent care from pregnancy through early childhood.

// education B.Tech Computer Science, Gautam Buddha University (2028)
// focus Humanoid perception · behavior systems · on-device ML
// based in India · open to research & engineering roles

02 Experience

AI Engineer

Sep 2026 – Present
iHub Research & Robotics Pvt Ltd
  • Developing Vision-Language-Action (VLA) systems for the TARA humanoid platforms, extending the perception and behavior stack built as an intern.
  • Contributing to research-paper publications alongside the R&D team.
VLAHumanoidPerception ROS2Research

AI Intern, R&D Department

Jul 2025 – Aug 2026
iHub Research & Robotics Pvt Ltd
  • Built humanoid perception pipelines with DINOv3, SAM 2.1, Depth Anything, and DUSt3R for object understanding, spatial reasoning, affordance detection, and scene segmentation.
  • Developed a Speech → Intent → Action pipeline integrating Whisper, VAD filtering, and an instruction-following LLM for command grounding in robotic tasks.
  • Implemented a VLA-driven behavior logic engine enabling reasoning-based task planning for healthcare (TARA Care) and education robots (TARA Edu).
  • Deployed AI models on Jetson Orin Nano with Dockerized pipelines, optimizing inference latency for real-world robotics constraints.
  • Prototyped TARA Care (patient-care humanoid) and TARA Edu (interactive educational humanoid) with the R&D team.
  • Partnered with TVS Automobiles on alloy defect & paint anomaly detection systems.
Jetson Orin NanoVLAVLM GPT-OSSWhisperDINOv3 SAM 2.1DUSt3RDockerROS2

Software Developer Intern

Dec 2024 – Present
Gautam Buddha University, Dean of Student Affairs
  • Building the GBU Hostels app and tech solutions for the DSA department.
  • Built data-driven dashboards and automation tools for administration.

Android Developer Lead

Jun 2025 – Aug 2025
Gautam Buddha University
  • Led development of the MyGBU Android app for streamlining student services.
  • Directed the technical team and shipped campus-focused solutions.

Community Speaker

Mar 2022 – Feb 2025
Arduino
  • Featured and interviewed by Arduino for outstanding projects.
  • Delivered talks on AI & IoT innovation and ran electronics workshops.

03 Selected Work

Industrial Deployment TVS Automobiles & WIL · Automotive Manufacturing

Humanoid Alloy Defect & Paint Anomaly Detection

Designed and deployed an industrial humanoid vision system for alloy defect and paint anomaly detection on a live automotive line. Building perception robust to harsh lighting, factory noise, motion, and clutter, operating well beyond controlled datasets.

  • Robustness under factory noise & harsh lighting
  • Real-time processing with motion-blur compensation
  • Deployed in a real manufacturing environment
  • Integrated as part of a humanoid perception pipeline
IEEE Publication GangaDrishti · IEEE IC3ECSBHI 2026 · Conference Paper

Real-Time Multi-Modal Sensor for Field-Based Microplastic Detection

A buoy-style field sensor fusing three modalities in one laminar-flow microfluidic cell: multispectral optics (400–940 nm LED array), impedance spectroscopy (1 kHz–2 MHz), and 500× microscopy on an ESP32-S3. A YOLOv8-C (CSPDarknet53) segmentation model feeds an attention-weighted fusion network predicting polymer class (PET, PP, PE, PS, PVC), size, and a contamination score — 8-bit TFLite at under 200 ms per sample, benchmarked against FTIR/Raman, with a companion app for geotagged contamination heatmaps and a Microplastic Risk Index.

Sensor FusionYOLOv8-CImpedance SpectroscopyESP32-S3TFLiteEdge ML
Read on IEEE Xplore
🤱

MaaKosh

AI-powered maternal & newborn wearable and app for pregnancy and newborn monitoring: fetal tracking, postpartum-hemorrhage and temperature sensing, and jaundice detection using embedded sensors and on-device ML. Devices tested at RML Hospital, Delhi.

Embedded MLWearablesHealthTech
Visit
🩺

VitaSense

Offline-first Android screener for ASHA workers (children 0–6): Green/Amber/Red triage into the RBSK referral pathway, synced to a hospital dashboard, with camera oculomotor tasks and a vitamin-D risk model fusing AS7341 spectral reflectance, pallor ML, and an XGBoost questionnaire. India Top 16, BioMed Bharat 2026.

Computer VisionXGBoostSpectral SensingBiomedical AI
Visit
⚙️

vla.cpp · Open Source

Two merged PRs to vla.cpp, a C++ inference runtime for Vision-Language-Action models (SmolVLA, π0, GR00T N1) built on llama.cpp for Jetson-class deployment: #12 — the Windows (WSL2 + CUDA 12.6) build path, validated with SmolVLA on an RTX 4050; and #17 — end-to-end LIBERO evaluation on WSL2 (vla-server lifecycle, MuJoCo EGL GPU access, uv cross-filesystem fixes).

C++CUDAllama.cppLIBEROVLA
GitHub
🏗️

FoundryBuild

Your AI CTO: turns an idea into an engineering blueprint via an adaptive CTO interview and an 11-agent pipeline (~13k LOC). Eight specialist agents run in parallel on FastAPI/asyncio with structured Gemini output, followed by a devil's-advocate critique and blueprint synthesis — a 9-tab plan covering architecture, costs, risks, and feasibility.

Next.js 16React 19FastAPIGeminiMulti-Agent
Visit
🧠

SecondMind

ESP32-S3 AI pendant that converts unintentional speech into structured tasks, summaries, and insights: STT + LLM processing + a vector DB for semantic memory recall. 2nd Place hackathon winner.

ESP32-S3STTVector DB
🗣️

BoloNyay

Multilingual legal AI system on Microsoft Azure + Bhashini: speech recognition, translation, legal NLP, and conversational assistance for public legal support. Govt. Top 100 Finalist, India.

AzureBhashiniLegal NLP
🌱

SkitterzAI

Plant disease detection and fertilizer recommendation AI for smart agriculture, bringing computer vision to the field for accessible crop diagnostics.

Computer VisionAgriTech
🌦️

VaPr Weather Station

A smart-farming weather station providing localized, sensor-driven insights to support better agricultural decisions.

IoTSensors
🤖

TARA Care & TARA Edu

Prototyped humanoid platforms (a patient-care humanoid and an interactive educational humanoid) with VLA-driven, reasoning-based behavior planning.

HumanoidVLAROS2

04 Skills & Stack

AI & ML

VLAVLMYOLOv8 CNNsTransformersWhisper VADTensorFlowOpenCV DINOv3SAM 2.1Depth Anything DUSt3RGPT-OSSLLaVA-style

Robotics & Perception

ROS2rclcppNavigation Multimodal FusionSpeech→Intent→Action SLAMBehavior Trees

Hardware & Systems

Jetson AGXJetson Orin NanoBlackwell STM32ESP32-S3Arduino Raspberry PiSensor FusionDockerLinux

Software & Cloud

PythonC / C++Kotlin / Java TypeScriptFastAPIFlask Next.js / ReactESP-IDFGemini API Multi-Agent OrchestrationSupabaseFirebase AzureBhashiniBackblaze B2Wasabi

Achievements & Recognition

🏅 NASSCOM AI CoE: Top 25 Selected Startups in India (with Govt. of Gujarat)
🚀 Selected: UP Deeptech Mission + BlueHill VC Pitching Round
🧬 India Top 16: BioMed Bharat 2026 Biomedical Device Hackathon
🌐 Qualified: Google's AI for Future Fund
🎤 Arduino.cc Community Speaker: AI & IoT innovation
📚 IEEE Paper Published: IC3ECSBHI 2026
🏆 India Top 100: Bhasha Bandhu (Microsoft × Bhashini)
🥈 Ignition Hackathon: 2× 2nd Place (2025, 2026)
🥇 Ignition IoT Challenge: 1st Place
🥈 GBU Entrepreneur Hackathon: 2nd Place
🧠 Science & Innovation Award: Amity University
🔬 Winner: CBSE Regional Science Exhibition 2023

Latest from X

Follow @_Vaastavikata on X

05 Contact

// let's build something that ships

Open to research & engineering roles, collaborations, and good problems.

[email protected] LinkedIn GitHub X