ForgeZen (Startup)

• Architected a provider-agnostic multi-agent AI platform (naming/brand generation, market intelligence, enterprise assessment) decoupled from any single LLM vendor via a providers/ abstraction layer.
• Designed model routing and cost-governance architecture tiering frontier vs. lightweight models to optimize cost vs. accuracy.
• Built a prompt registry and versioning system with Pydantic-validated, auditable structured outputs.
• Implemented RAG grounding with citation display over a 210-document enterprise knowledge base using pgvector.
• Set multi-tenant SaaS security architecture: tenant isolation, RBAC, per-tenant AI cost tracking, and audit logging.
• Owned full-stack technical direction (Django/DRF, Next.js, Celery, Azure) as sole technical decision-maker.

CohnReznick LLC

• Reported to CohnReznick's Global Head of AI, designing and deploying AI-native solutions for financial, banking, and auditing workflows.
• Architected AI-enabled applications using LLMs, SLMs, RAG/GraphRAG, and NLP for enterprise workflow transformation.
• Built full-stack applications integrated with Claude Code, GitHub Copilot, and Azure AI services.
• Designed secure data connectors and integrations with cybersecurity, risk, and IAM teams.
• Led company-wide AI initiatives, aligning delivery with leadership and business objectives.

Handshake AI — Project Proctor (Project Darwin)

• Training LLMs to better support STEM-based reasoning.

Mercor (contracted to a leading AI research lab)

• Orchestrated reproducible ML pipelines and LLM training workflows under ephemeral execution constraints for agentic-trajectory generation.
• Directed LLM-guided EDA strategies across Kaggle-style datasets with GPU-optimized PyTorch DataLoaders.
• Architected prompt-driven orchestration for Slurm-scheduled NVIDIA A100 GPU environments on Rocky Linux.
• Applied gradient checkpointing, mixed-precision (FP16/BF16), and distributed data parallelism to maximize A100 utilization.
• Shaped LLM-agent problem-solving trajectories end-to-end, from EDA to full-dataset optimization, for autonomous AI research.

VerzX / Zintro

• System reliability & risk assessment; MLOps for edge computing, industrial IoT, and large-scale IoT device deployments.

VisasQ / Coleman

• AI/ML solution deployments and project lifecycle management.

GIS Ltd

• Architected secure, data-privacy-first digital transformation templates for large and medium enterprises.
• Defined data engineering, management, architecture, privacy, and cybersecurity specifications for enterprise and government digital transformation.
• Evaluated Google ADK, Vertex AI, Google AI Studio, Workspace, and Gemini for enterprise AI transformation.

Corning Incorporated — Painted Post, NY

• Generative AI (2023–2024): Built a customized ChatGPT interface for industrial use with Docker, PostgreSQL, Vue, and Django; enabled RAG on GCP Vertex AI with GPT4All.
• Software Engineer (2023): Built anomaly detection for glass-surface fault contours using Databricks (Delta Lake, Unity Catalog), Spark, PyOD; shipped a Tkinter/PyInstaller desktop app that cut product waste 31%.
• Lead Architect, Computer Vision & IoT (2023): Deployed real-time vision inspection on Jetson AGX Orin at <15ms latency / 99.2% accuracy, cutting defect escape rate 40%; led MLOps CI/CD to 50+ edge devices via AWS SageMaker + IoT Greengrass.
• NLP Engineer (2022–2023): Built Spacy/Gensim/TF-IDF/LSI + LSTM fault-prediction models on Kafka/Flink streaming data, cutting fault ID time ~97%; deployed on GCP Cloud Run with BigQuery ML.
• ML Engineer (2020): Built a VGG16 + LLM-prompted image recommendation system with FAISS/HNSW/IVF nearest-neighbor search, cutting fault inferencing from ~2 weeks to <1 minute.
• Project Engineer, IoT Vision (2020): Led a YOLO + Jetson Nano/Xavier COVID-19 social-distancing system using GeoPandas/Fiona/Shapely for geofencing; 93% cheaper than the comparable vendor solution. Also built a vaccine-vial tracking system with Hough Transforms and CUDA-accelerated OpenCV.
• Project Engineer, Smart Manufacturing (2019–2023): Led a cross-functional team (Corning, Verizon, AWS, Gestalt Robotics) defining private-5G AGV/AMR enterprise architecture; holds a resulting US Patent and an IEEE-published journal paper with Verizon engineers.

Comcast — Philadelphia, PA

• Analyzed real-time and batch data via Databricks, Presto, Spark, Kafka, and AWS (Kinesis, S3, Elasticsearch, Lambda).
• Analyzed network health data across ~29M devices plus clickstream data for root-cause analysis, improving customer experience and reducing churn.

Drexel University — Philadelphia, PA

• Taught applied/computational physics with emphasis on big-data technologies; consistently scored high student evaluation ratings.

Ph.D. & M.Sc. Studies — Drexel University / Blekinge Institute of Technology

• Ph.D. focused on embedded wavelet-based statistical ML at the edge of large IIoT/smart-grid networks using a fixed-point C28x DSP.
• Developed the industry's first Embedded-C wavelet suite for the C28x DSP, applied to ECG signal characterization and low-voltage powerline analysis.
• Worked briefly with the US Department of Energy studying grid effects of large non-linear loads (EVs, data centers) and mixed renewable/non-renewable generation.
• Applied QGIS/PostGIS spatial risk assessment to model weather impact on grid stability.

Rechez Communications — Awka, Nigeria

• Designed an enterprise-grade network data and monitoring architecture across LAN, WAN, VSAT, and consumer access networks on Fedora Linux.
• Led the technical strategy behind NCC regulatory license approval, enabling expansion into underserved regions.
• Redesigned vendor/connectivity strategy, cutting recurring operational costs by 42%.