Work · builds

Experience and projects

Selected roles and builds. Shipping AI systems, research, and software with customers and labs.

AI systemsRAGVLMsFull-stackSecurity

Experience

Newest first. Roles that shaped how I ship.

Aug 2026 – Present

Associate Forward Deployed Engineer

Momentuum Blue, Coforge

Shipping AI systems with customers at the edge of production.

DeploymentAI systems
May 2025 – Apr 2026

Graduate Student Research Volunteer

UF Data Studio

  • Built a Python MLSecOps evaluation framework with Docker and GitHub Actions for 50+ weekly benchmark runs across 8 VLMs and 7 object detectors.
  • Quantified domain gaps across VLMs and detectors against natural-image baselines.
  • Benchmarked 6 VLMs for accessibility and adversarial robustness.
PyTorchVLMsTrustworthy ML
Sept 2023 – Jul 2024

Software Development Engineer

Simpel Techlabs

  • Architected an offline-capable React/Node.js/IndexedDB PWA on AWS serving 100,000+ rural schools; reduced origin load by 65%.
  • Reduced latency by 20% and P99 by 35% using Redis, CloudFront, and ELB tuning across four AWS regions.
ReactNode.jsAWSIndexedDB
Sept 2022 – Aug 2023

Compliance Analyst

TresVista

  • Automated security-alert workflows with Power Automate and APIs, reducing mean response time by 60%; led SOC 2, ISO 27001, and GDPR rollout and mitigated 70% of VAPT risks.
Power AutomateSecurityVAPT
Oct 2018 – Mar 2019

Cyber Security Intern

CloudOptics.io

  • Assisted with VAPT scans on client networks and automated client-asset assessments.
  • Helped develop a Cyber Exposure Index for monitoring threat levels.
VAPTExposure Index

Education

Learning across systems, research, and engineering.

University of Florida

Master of Science, Computer Science

Coursework:Adv Data Structures · ML Engineering · NLP · Trustworthy ML

May 2026

RV College of Engineering

Bachelor of Engineering, Electrical and Electronics

Coursework:Digital Signal Processing · Discrete and Integral Transforms · Modern Control Theory

Feb 2023

Projects

Selected research and side builds.

01 · RAG

Doofenshmirtz Evil Inc RAG

Local RAG over policy docs with Ollama embeddings/generation, Chroma retrieval, Cohere, and an evaluation harness for recall and answer accuracy.

Python · Ollama · Chroma · Cohere · RAG
02 · MCP

Claims MCP

Model Context Protocol server for claims workflows. Early tooling for agent-ready integrations.

MCP · Agents
03 · Inference

Energy-Efficient LLM Inference Pipeline

8-bit DistilBERT quantization (~74% lower latency, ~65% less energy on NVIDIA T4), TextAttack adversarial training, FastAPI + Docker deploy, MLflow and CodeCarbon tracking.

PyTorch · Hugging Face · ONNX · TextAttack · FastAPI · Docker · CodeCarbon
04 · Agents

Autonomous Multi-Agent Energy Futures Monitor

CrewAI multi-agent daily energy and commodities reports with forecasting, technical analysis, and news aggregation; Streamlit caching cut report time ~30%.

CrewAI · LLM Agents · LSTM · Streamlit · Google Cloud
05 · Research

VLM Hallucinations in Comics: Why Semantic Similarity Fails

Evaluated 6 open-source VLMs on a 150-page comic/manga set; showed why cosine similarity and KL divergence miss object, description, relation, and attribute hallucinations.

Vision-Language Models · Evaluation · Hallucinations · Multimodal ML
06 · NLP

Sentiment-Aware Question Answering

DistilBERT/Hugging Face QA paired with sentiment analysis and a Streamlit UI for sentiment-tailored responses.

Python · NLP · QA · Sentiment Analysis