themeAcademicCorporateTerminalMemeWrappedRoomGame✕

available/open to opportunities · Canada & remote

Hi, I'm Dvip.

I build production systems

AI Engineer — NLP & Intelligent Systems. I ship end-to-end intelligent systems — from LLM pipelines and retrieval stacks to production Laravel migrations — with a bias toward trustworthy, auditable design.

Oshawa, Ontario, Canada
view as

tailor this page to the role you're hiring for

~/dvip/system.maplive
Your problemmessy · real-worldNLP + LLM pipelinesTRAC-RE · 1 paperAutonomous agentstool use · guardrailsCloud infraLambda · EC2 · DockerProduction systems85–90% auto-resolvedShippedthings I've built
▸hover or tap a node to jump to it
impact.readout
experience
0 yrs
y1
y2
y3
y4

Building production systems end-to-end

auto_resolution
0–90%
085–90

Autonomous migration issue resolution

manual_effort
0%+
before
after

Reduction in manual migration effort

cost_per_run
<$0
$0cap $80 / run

Cost per autonomous migration run

Infrastructure

Cold Start Race

A live benchmark: four AWS deployment strategies run the exact same workload head-to-head and stream their real millisecond timings back as they finish. Hit start and watch actual infrastructure race — no mocks.

AWS LambdaEC2 + DockerSSE streamingRedis rate limitingEdge concurrency caps
~/race/lanesready

$ same workload, four setups, one start signal. Faint markers show typical times from past races.

  • Lambda Cold Start

    AWS Lambda with forced cold start — no warm container reuse

    typical~1.2–1.5 s

  • Lambda Warm

    AWS Lambda kept warm via EventBridge ping every 5 minutes

    typical~37 ms

  • EC2 Direct

    Fastify on bare EC2 (t3.micro) behind Nginx — port 3001

    typical~90 ms

  • Docker on EC2

    Containerized Fastify via docker-compose on EC2 — port 3002

    typical~95 ms

Races all 4 strategies simultaneously — typically 1–3 seconds

~/race/cost.model

What would this cost at your traffic?

The race times barely change between runs, so the interesting number is the bill. Set a traffic level and see what each setup costs per month at current AWS prices.

traffic
100Krequests / day
lambda memory
billed duration per requesttypical race result
ec2 instances
1 × t3.micro

▸ Warm Lambda is cheaper than one t3.micro until about 1.2M requests/day; with every request a cold start, that drops to 81K.

Docker on EC2 costs the same as EC2 direct: same host, same bill. Its price is paid in latency, not dollars.

AI Demo

Autonomous Agent

A real AI agent you can run. Pick a task, give it an input, and watch it reason, choose tools, call real APIs, and write a structured report — every step streamed live as it happens.

Tool-use loopSSE streamingRedis rate limitingCost ceilingOutput sanitizationOpenRouter fallback
~/agent/select-taskready

$ pick a task and give it an input, or hit run example to watch it work right away

15 tools
2 tools
3 tools

Experience

Where I've worked

~/experience/deploys.log4 deploys · 1 running

  1. deploy #03Mediabridgestatus: running

    Founding Engineer

    Mediabridge · Remote

    Apr 2025 — PresentPart-timeindustry

    • Co-built and launched a multi-tenant SaaS platform from scratch, owning architecture decisions across frontend, backend, and AWS deployment.

    • Personally built the core systems — a custom role-based access control (RBAC) layer, a dynamic form builder, and an event-driven notification system.

  2. deploy #04Palomino Systemsstatus: shipped

    AI Engineer

    Palomino Systems · Remote

    Nov 2025 — Jun 2026Part-timeindustry

    • Designed and shipped Laravel Upgrader, a fully autonomous CLI-driven AI agent that migrates legacy Laravel codebases end-to-end — a 10-step pipeline (analysis → planning → transformation → validation → self-healing) processing ~10k LOC/run at under $80/run, with 85–90% automated issue resolution via detect → fix → retry loops.

    • Measured two full migrations by hand first, then automated those exact steps — cutting manual migration effort by ~80% (a baseline-grounded number, not an estimate).

  3. deploy #02Ontario Tech Universitystatus: shipped

    Graduate Research Assistant (Responsible AI & Evals)

    Ontario Tech University · Oshawa, ON

    Sep 2024 — May 2026research

    • Designed TRAC-RE — a Traceable, Reliable, Auditable, and Contextual framework for automated Requirements Engineering, built from four interconnected phases that take raw stakeholder prose to implicit domain-knowledge discovery, grounded in an industrial FinTech/SaaS corpus of 110 requirements and 1,997 High-Level JSON (HLJ) artifacts.

    • Phase 1 — Context-aware requirement identification: empirical study of sentence-level classification (all-mpnet-base-v2, DeBERTa-v3-base) across frozen / LoRA / full fine-tuning and context windows k=0–3, showing structured local context adds +16 F1 points (0.664 → 0.894) and that 4K domain-aligned samples (F1=0.894) beat 15K mixed (F1=0.883).

  4. deploy #01Finserve Infotechstatus: shipped

    Full-Stack Developer

    Finserve Infotech · Remote

    Jan 2024 — Dec 2024industry

    • Delivered 4 production systems for clients across three industries (manufacturing, retail distribution, travel), building each full stack in Laravel and MySQL, from UI and application logic to the transactional schemas underneath.

    • Built a manufacturing execution system tracking garment production end to end, from raw cotton to finished product, including routing work out to external processors and back, with bidirectional invoicing covering both standard and reverse invoice flows for subcontracted stages.

Research

Publications & thesis

ThesisMASc, Software Engineering · Ontario Tech University

TRAC-RE

Toward Automated Requirements Engineering: Empirical and Architectural Foundations for Structured Parsing and Knowledge Discovery

TRAC-RE: Traceable, Reliable, Auditable, and Contextual

TRAC-RE is a framework for automated requirements engineering (RE) built from four interconnected phases. Together they take raw stakeholder prose to implicit domain-knowledge discovery — spanning context-aware requirement identification, audit-grade keyword extraction, governed LLM parsing into structured JSON, and multi-signal implicit keyword discovery, all grounded in an industrial FinTech and SaaS corpus.

Supervisor: Prof. Sanaa Alwidian

Traceable
Every keyword, tag, and classification carries per-decision provenance back to its source artifact.
Reliable
Versioned governance and confidence scoring make pipeline outputs reproducible and defensible.
Auditable
A bounded LLM-as-judge design and SME-audited samples keep human verification in the loop.
Contextual
Local-context modeling, not raw model scale, drives requirement identification accuracy.

Papers

01

  1. PublishedCASCON 2025 · 2025

    Improving Reliability of LLMs in RE with Structured Confidence & Tag Governance

    A modular multi-model LLM pipeline that converts raw stakeholder requirements into High-Level JSON (HLJ) artifacts with confidence-scored fields, paired with a versioned tag-governance system that catches prompt-leak exploitation, hallucinated tags, and low-agreement outputs before they reach downstream stages.

    • LLM governance
    • Requirements Engineering
    • Tag validation
    Read → Improving Reliability of LLMs in RE with Structured Confidence & Tag Governance (opens in a new tab)

Want the full research write-up? Switch to the academic view →

Capabilities

What I work with

  • Research Areas

    • Requirements Engineering
    • NLP for Software Engineering
    • LLM Governance
    • Knowledge Discovery
    • Empirical SE

  • Languages

    • Python
    • TypeScript / JavaScript
    • PHP (Laravel)
    • SQL
    • Bash

  • AI / ML

    • PyTorch
    • HuggingFace Transformers
    • sentence-transformers
    • SBERT / MPNet
    • MiniLM
    • DeBERTa-v3
    • KeyBERT
    • RAKE
    • FAISS
    • UMAP + HDBSCAN
    • LoRA Fine-tuning
    • RAG Pipelines
    • Anthropic / Claude API
    • MCP
    • Aider

  • Frontend

    • React
    • Next.js
    • Tailwind CSS
    • Zustand

  • Backend / Cloud

    • Laravel
    • Flask
    • REST APIs
    • Microservices
    • RBAC
    • Docker
    • AWS
    • CI/CD
    • GitHub Actions

  • Testing

    • Vitest
    • fast-check (property-based)
    • Testing Library

Education

  • MASc, Software Engineering (Thesis)
    Ontario Tech University · Oshawa, ON, Canada
    2024 – 2026

Certifications

  • AWS Certified Cloud Practitioner
    Amazon Web Services · 2024

Teaching & Authored

  • OVIN EV Micro-Credential Course
    Ontario Vehicle Innovation Network · 2024

Featured work

Things I've built

Two systems tell the story best: a production migration platform that pays for itself, and the research pipeline behind my thesis. Here's what each one does and what it's made of.

  1. Product

    Inbox Intelligence Platform

    Gmail-native communication intelligence layer — one ingestion pipeline, six analysis modules.

    Built at a 48-hour hackathon. Turns unstructured email into structured, actionable intelligence: a single ingestion + parsing core (Claude classifier, entity extractor, temporal indexer) feeds six independent modules — financial waste detection, relationship decay signals, contract/obligation tracking, follow-up commitments, a RAG-based institutional memory, and health-admin tracking. One parse, many consumers — no redundant API calls, and an entity graph that compounds over time. Currently being taken from hackathon build to a live product.

    What it's built with

    1. Claude Sonnet
    2. Gmail MCP
    3. Google Calendar MCP
    4. RAG / pgvector
    5. PostgreSQL
    6. React
    7. Tailwind
    8. Node.js
  2. Product

    Laravel Upgrader

    Autonomous 10-step AI migration agent for legacy Laravel codebases.

    CLI-driven migration agent that ingests legacy Laravel code and runs analysis → planning → transformation → validation → self-healing loops. Processes ~10k LOC per run at under $80 cost, with 85–90% automated issue resolution.

    What it's built with

    1. Python
    2. Claude / LLM Orchestration
    3. Laravel
    4. CLI
    5. Self-healing retry loops
  3. Research

    TRAC-RE — Automated Requirements Engineering Framework

    A Traceable, Reliable, Auditable, and Contextual framework for automated Requirements Engineering — structured parsing, context-aware classification, and implicit keyword discovery.

    The research framework behind my MSc thesis at Ontario Tech University. TRAC-RE — Traceable, Reliable, Auditable, and Contextual — converts raw stakeholder prose into audit-grade knowledge artifacts across four interconnected phases:

    1. 1.Phase 1 — Context-aware requirement identification — Sentence-level classification with structured local-context features. F1 improved 0.664 → 0.894 on 110 FinTech/SaaS documents; 4K domain-aligned samples (F1=0.894) beat 15K mixed (F1=0.883).
    2. 2.Phase 2 — Audit-grade keyword extraction — Perfect-precision extraction that surfaced a structural coverage ceiling of 1.5 canonical keywords per artifact, motivating the implicit-discovery phase.
    3. 3.Phase 3 — Governed LLM parsing (HLJ) — Multi-model parsing (GPT-4.1, Claude Opus 4, Meta-70B) into confidence-scored High-Level JSON with versioned tag governance (v0 → v2); tag precision 0.657 → 0.897. Supporting paper published at CASCON 2025.
    4. 4.Phase 4 — Implicit keyword discovery — A 5-stage engine over a 275,164-edge enriched co-occurrence graph, with UMAP+HDBSCAN cluster-gap detection and a bounded LLM-as-judge tiebreaker.

    Cross-cutting infrastructure — 1,997 HLJ artifacts, a 13,725-entry domain dictionary, a FAISS index over 768-dim SBERT vectors, full per-decision audit logging, and drift monitoring across pipeline runs. The framework's central argument: governance and distributional alignment, not raw model scale, are the primary levers for trustworthy RE automation.

    What it's built with

    1. Python
    2. PyTorch
    3. HuggingFace
    4. SBERT / MPNet
    5. DeBERTa-v3
    6. FAISS
    7. UMAP
    8. HDBSCAN
    9. LoRA

Also on this page — live & playable

Infrastructure↑ Run it

Cold Start Race

Four AWS deployment strategies benchmarked head-to-head with live millisecond timings.

AWS LambdaEC2 + DockerSSE streamingRedis

AI Demo↑ Run it

Autonomous Agent

A real tool-using AI agent that reasons, calls live APIs, and streams a structured report.

Tool-use loopSSE streamingRedisCost ceiling

Open source

GitHub at a glance

@dvip07 →
Public repos
28
Total stars
1
Followers
9
Top language
Python

Contact

Let's build something

Have an AI or engineering problem worth chasing?

Happy to chat about NLP pipelines, LLM governance, autonomous migration systems, or research collaboration. Email is the fastest way through.

✉ dvippatel.math@gmail.comLinkedInGitHub