Suujay Dhhoka, AI Engineer — Agentic AI, RAG & Robotics

About me

Engineering
intelligence into
the real world

I'm an AI Engineer with an MS ECE from Rutgers University (GPA 3.50), building agentic AI systems, Retrieval-Augmented Generation (RAG) pipelines, and multi-agent orchestration for fintech and enterprise workflows. My work spans LLM APIs, vector retrieval, and cloud-native ML infrastructure on AWS and Azure, with a robotics and computer-vision background underneath it.

At Kaizen Studio, I built AWS Lambda preprocessing plugins and migrated batch inference to autoscaling SageMaker endpoints. As a Teaching Assistant at Rutgers, I mentored 550+ students across Machine Vision, Cloud Computing, and Digital Signal Processing.

I'm fluent in English and Hindi, with working German and elementary Spanish, and experience across research labs, cloud companies, and creative-tech studios. Based in Edison, NJ and open to full-time roles in AI engineering, applied AI, and MLOps, particularly in fintech and enterprise AI.

Graduate GPA

3.50

Students mentored

550+

Published research papers

3

Cloud certifications

2×

Major projects shipped

9+

Expertise

Technical skills

01

Generative & Agentic AI

RAGHybrid RAG (BM25+Vector+RRF)Multi-Agent SystemsMCPClaude APIOpenAI APILangChainPrompt Engineering & CachingFAISSChroma

02

Machine Learning & MLOps

PyTorchTensorFlowModel Fine-TuningCNNsReinforcement LearningAnomaly DetectionExperiment TrackingModel Monitoringscikit-learn

03

Data & Cloud

AWS (SageMaker, Bedrock, Lambda, S3, EC2)Azure (Data Factory, AKS, DevOps)GCP Vertex AIDockerKubernetesSparkSnowflakeMLflowCI/CD

04

Programming

PythonJavaC++SQLBashJavaScriptReact.jsFastAPIStreamlit

05

Robotics & Perception

ROS / ROS2SLAMGazeboMuJoCoSensor FusionLiDARForce ControlHRI

06

Computer Vision

OpenCVYOLOv8GStreamerObject DetectionSegmentationReal-time CV

Work history

Where I've worked

Kaizen Studio LLP

Jun 2025 – Dec 2025

Internship

Machine Learning Engineer

AWS SageMakerPyTorchTensorFlowOpenCVS3 · Lambda · EC2

Built Python/AWS Lambda preprocessing plugins that classified and tagged incoming assets via a CNN classifier, improving throughput.
Migrated batch inference to autoscaling AWS SageMaker endpoints with MLflow experiment tracking, cutting manual steps by 40%.
Fine-tuned PyTorch image-classification models for automated visual QA, reducing manual review workload by 25%.
Set up a GitHub Actions CI/CD pipeline with model-validation gates and unit tests, cutting post-deployment regression bugs by 20%.

Rutgers University

Jan 2024 – May 2026

Academic

Graduate Teaching Assistant

Machine VisionCloud ComputingComputer ArchitectureDSP

Redesigned lab materials and led weekly office hours for 550+ students across 4 courses, raising project pass rates by 30%.
Reviewed 40+ ML/CV capstone projects per semester (OpenCV, PyTorch, TensorFlow, AWS/Azure), providing direct design feedback.
Built Python autograding scripts for programming assignments, cutting grading turnaround time for the teaching staff by 50%.
Authored supplemental lab guides and starter notebooks later adopted as standard course material across 2 subsequent semesters.

IBN Technologies Ltd.

Jan 2023 – Dec 2023

Full-time

Cloud Engineer (promoted from Internal Tech Support)

Azure Data FactoryAzure AKSSnowflakeDevOpsAzure Monitor

Built Azure Data Factory ingestion pipelines feeding Snowflake, cutting nightly batch processing by 35% for analytics and ML features.
Deployed containerized ML inference as FastAPI endpoints on Azure Kubernetes Service (AKS), sustaining 95% uptime.
Configured model-performance dashboards and drift-detection alerts in Azure Monitor, cutting incident-detection time to under 15 min.
Scripted Azure DevOps CI/CD pipelines with automated model-validation checks, reducing release turnaround time by 20%.
Implemented IAM role-based access and secret rotation across cloud resources, closing 12 findings in a security audit.

IBN Technologies Ltd.

Jul 2022 – Dec 2022

Full-time

Internal Tech Support

Internal Tech SupportIT Operations

Provided internal technical support before being promoted into the Cloud Engineer role.

VIIT, Research Lab

Oct 2022 – Dec 2022

Research

Robotics Research Intern

ROSPythonReinforcement LearningComputer Vision

Implemented ROS-based control algorithms for a robotic arm, improving motion accuracy and repeatability.
Integrated ML-based object detection for grasp optimization; conducted RL experiments for motion fine-tuning.

Portfolio

Things I've built

Agentic AI · Fintech · Featured

TradeDesk: Multi-Subagent Equity Research System

A pipeline of 5 parallel specialized subagents (fundamentals, technicals, sentiment, macro, risk) plus a synthesis agent using extended thinking, cutting research turnaround to under 5 minutes. Backtested agent-generated strategies against historical price data, benchmarking theses against simple buy-and-hold baselines.

Multi-AgentStreamlitClaude APIAWS BedrockExtended Thinking

View on GitHub ↗

Agentic AI · Fintech · Featured

ReconcileAgent: Autonomous Transaction Reconciliation

A two-pass deterministic matching engine with a durable SQLite audit trail and human approval gate, auto-resolving most transactions without a model call. Cut inference cost by roughly 60% versus an all-Sonnet baseline via tiered Haiku/Sonnet routing keyed to match difficulty.

Claude APIMCPSQLiteHaiku + SonnetPandas

View on GitHub ↗

Agentic AI · Fintech · Featured

FinGuard: Agentic Regulatory Compliance Audit System

Hybrid RAG retrieval (BM25 + FAISS vector search, fused via Reciprocal Rank Fusion) over regulatory text through MCP tools, improving passage relevance over single-method search. Added a two-stage LLM-as-judge evaluation framework and prompt caching to catch hallucinated citations and cut latency.

Hybrid RAGMCPExtended Thinkingsentence-transformers

View on GitHub ↗

AI/ML · Featured · GitHub

Quantum-Enhanced Continual Learning Fabric

Federated continual learning system for planetary climate forecasting across 3 geographic nodes. Combines FedAvg, Elastic Weight Consolidation, Physics-Informed Neural Networks, and Multi-Agent PPO. Ships with 29 tests and a full CI/CD pipeline.

PyTorchFedAvgPINNMulti-Agent PPOEWCCI/CD

View on GitHub ↗

AI/ML · Featured · GitHub

Job Success Predictor

AWS-powered ML system that scores resume-to-job fit. Paste your resume, get match scores against 20 real ML engineering roles, and receive exact skills to add for a measurable boost in match rate.

AWSNLPResume ParsingML ScoringPython

View on GitHub ↗

Cloud · ML

Fraud Detection Pipeline

Explainability-first fraud system with XGBoost, SHAP, SMOTE, and MLflow, deployed on AWS SageMaker.

XGBoostSHAPSageMakerMLflow

NLP · MLOps

Domain-Adaptive Classifier

Fine-tuned transformer with RAG, MLflow tracking, and FastAPI deployment on AWS S3.

RAGHugging FaceFastAPIDocker

RL · Cloud

Warehouse Multi-Agent RL

Multi-agent RL logistics optimizer using RLlib, PPO, and Gymnasium on GCP Vertex AI.

RLlibPPOGCP Vertex AI

Robotics · HRI

Campus Guide Robot

LLM-driven campus robot with SLAM navigation, YOLOv8 perception, and MuJoCo simulation.

SLAMLLMsYOLOv8MuJoCo

Robotics · Research

Vision-Guided Scrubbing Robot (TEDSR)

Hybrid force-position controlled scrubbing robot with real-time YOLOv8 perception and ROS-based sensor fusion for compliant, high-precision surface manipulation.

ROSYOLOv8PyTorchForce ControlOpenCV

Robotics · Research

Multi-Robot Search & Rescue System

Autonomous multi-robot SAR platform using SLAM, deep RL exploration, YOLOv8 victim detection, and swarm coordination on Jetson Xavier NX hardware.

ROS 2GazeboSLAMDeep RLLiDAR

Academic background