Portfolio · 2026
Dhawal
Gajwe
AI Engineer. Building intelligent systems — one algorithm at a time.
Approach
Ideas used to wait.
Now they ship.
01 — About
Building with AImade me more creative.
I'm an AI Engineer with an MS in Information Science (Machine Learning) from the University of Arizona. I work across transformers, generative AI, and context-aware systems — shipping production ML in Python and obsessing over the architecture decisions that turn code into a product.
Languages
AI / ML
LLMs / Agents
Web
Data
Infra
Activity
02 — Selected Work
A few things I've had the privilege of building.
PubMed RAG
MS Capstone · Team project
A RAG system answering biomedical queries over 18,000 PubMed articles with cited PubMed IDs — LangChain ReAct + Weaviate.
Voice AI Order Agent — Digirestro
Client work · Built from scratch
Voice agents that take restaurant orders and reservations by phone, integrated with the client's kitchen POS. The AI was the easy part — real-time audio, barge-in, and accents are where the engineering lives.
ProdShoot — AI Product Photography
MVP · Demo below
Turn a simple product photo into professional product photography by chatting with AI — new backgrounds, props, and compositions, with selectable image-generation models.
Dhammapada Comics
Shipped · iOS App Store
All 423 verses of the Dhammapada as AI-illustrated comics — translations and the story behind each verse. Fully offline, no accounts, no ads.
LockedIn — Focus Timer
Shipped · iOS App Store
A minimalist focus timer with a retro LCD interface — block and Pomodoro sessions, plus a Strict Mode that blocks distracting apps. 100% private, on-device.
ColdConnect
Live · coldconnect.vercel.app
An AI tool that connects you with the right job recruiters — finds the contact, drafts the cold email, sends it for you.
Transformers from Scratch
Code + 4-part article series
A Transformer implemented from scratch in PyTorch for English→Italian translation — custom multi-head attention, positional encodings, training loop.
03 — Experience
Six years shipping production systems.
AVSI Systems
AI Engineer
Client: FedPointJun 2026 — Present
PORTSMOUTH, NH
- Authored a Terraform IaC stack of 80+ resources (VPC, IAM, KMS, Lambda, ECS, SQS, API Gateway, WAF, CloudWatch, Secrets Manager) that standardizes environment parity and cuts deployment time.
- Built a serverless RAG chatbot admin platform on AWS (API Gateway, Lambda, ECS Fargate, SQS, S3, OpenSearch) with a JWT-secured CloudFront + S3 admin UI — administrators deploy and update chatbots across environments in minutes.
- Designed a TypeScript/Fastify admin API with 60+ endpoints for multi-tenant management of applications, chatbots, environments, RBAC, and knowledge bases, on pluggable Postgres/SQLite adapters.
WEX
AI/ML/NLP Engineer
Aug 2025 — Feb 2026
CHICAGO, IL
- Designed and scaled an AI-driven claims auto-adjudication system (LangChain, OpenAI LLMs, OCR pipelines) — 40% of insurance claims reimbursed automatically with no manual intervention, served by FastAPI inference endpoints for real-time decisioning.
- Built end-to-end NLP pipelines extracting, normalizing, and validating structured data from unstructured claim documents, with MLOps practices for automated testing and monitoring.
- Clustered historical claims to find high-confidence adjudication patterns, expanding automated coverage by a further 2%; deployed models on AzureML.
McAfee
Software Engineer
Aug 2019 — Aug 2024
INDIA
- Led design of App Control plugin features using OSQuery and Auditd, improving endpoint threat detection across Linux/Unix environments.
- Implemented CI/CD pipelines with Jenkins and Groovy, cutting manual testing effort by 40%; standardized Kubernetes/Docker deployments, reducing operational errors by 30%.
- Strengthened CIS security hardening in Python, Lua, and C++, improving compliance scores by 25%.
IBM
Software Engineering Intern
Jun 2018 — Jul 2018
INDIA
- Built a web-based CMDB interface (Node.js, Python, MongoDB) to visualize dynamic configuration data for enterprise IT operations.
- Boosted data retrieval speeds by 25% through schema optimization and efficient indexing in MongoDB.
04 — Writing & Videos
I build in public — and explain how.
Build a 24/7 AI Dev Team: How to Deploy Autonomous Agents
Agents that write code and open PRs, running around the clock.
Understanding ReAct LLM Agents using LangGraph
How ReAct agents reason and act, built step by step in LangGraph.
I Used Claude Skills to Transform My Landing Page Design
Step-by-step redesign driven by Claude Skills.
ColdConnect — an AI tool that sends cold emails for you
Automating job outreach in one click.