hey, i'm Darshan.
ai engineer & software developer with 4+ years of experience building production-grade ai platforms, distributed systems, and full-stack applications. i work primarily with go, python, and typescript.
originally from surat, india, now based in austin, texas. i moved 9,000 miles to pursue a master's in computer science at ut arlington, and i've been building things at the intersection of ai and infrastructure ever since.
i believe in shipping fast, building systems that scale, and never waiting for permission to start something ambitious.
now
formerly sde at ftnss inc.
led backend services and RESTful APIs for a multi-tenant marketplace platform now live across 40+ countries, cutting average API response time by 45% and release cycle time by 40%
building Regent
production-grade saas that replaces human executive assistants with a 13-stage email processing pipeline, RAG-backed drafting, multi-model routing with circuit breakers, and 4-tier Stripe billing
building MailPilot
real-time email automation with Go async job workers, IMAP sync, cron scheduling, and SSE broker pushing live pipeline status
master of science in computer science
university of texas at arlington · ai systems, distributed computing, full-stack engineering
experience
ai engineer
Dec 2025 - Apr 202633 strategies llc · remote · austin, tx
delivered an ai-powered fitness platform (Plya) for a client: ingests short-form video (Instagram Reels, YouTube Shorts, TikTok) and auto-tags clips against user onboarding profiles to drive personalized content feeds and an ai coach feature.
- -integrated LLM APIs from the TypeScript/Node layer to auto-tag ingested clips against user onboarding profiles (goals, fitness level, Q&A responses) and power an in-app ai coach agent, with prompt engineering and iterative output evaluation for reliable tagging accuracy
- -developed the React and TypeScript frontend with REST API integration and Supabase real-time data, delivering responsive user-facing experiences and operational views across web clients
- -owned end-to-end deployment and release automation on Railway with CI/CD pipelines, environment management, and iterative performance tuning across staging and production, shipping client-facing features rapidly without regressions
software development engineer
Jun 2025 - Jul 2026ftnss inc. · remote · austin, tx
joined as lead sde on a multi-tenant 'airbnb for gyms' marketplace platform. architected backend services and RESTful APIs with normalized relational schemas supporting vendor onboarding, real-time booking workflows, and dynamic availability across web and mobile clients.
- -optimized PostgreSQL with composite indexing strategies, row-level security policies, and real-time subscriptions, cutting average API response time by 45% and enabling sub-100ms booking updates for all concurrent platform users
- -deployed production infrastructure on AWS using Terraform with auto-scaling groups, load balancers, health checks, and blue-green deployments
- -built end-to-end GitHub Actions CI/CD pipelines that reduced release cycle time by 40%
- -set coding standards and review processes for the engineering team, collaborating with product to turn business requirements into technical roadmaps and releases
software development engineer
Dec 2021 - Jun 2023acodesoft technologies llc · surat, india
started as an intern building backend modules in c# .net core, promoted to full-time sde based on consistent delivery. led migration from monolithic architecture to microservices.
- -led migration of monolithic .net services to java spring boot microservices on docker/kubernetes, reducing downtime by 35%
- -built high-throughput rest apis with mssql tuning, improving p95 query latency by 60%
- -implemented centralized oauth 2.0 / jwt authentication with rbac for inter-service communication
- -deployed high-availability environments on aws ec2 with auto scaling, maintaining 99.9% uptime
signal
6 recs“the rare engineer who thinks holistically and operates like an owner.”
Mbiyimoh Ghogomu · founder & ceo, 33 strategies llc
“the person i trusted to make it happen.”
Jordan Robinson · 2x founder & ceo, ftnss inc.
education
master of science in computer science
university of texas at arlington · usa
ai systems, distributed computing, full-stack engineering
bachelor of technology in computer engineering
p p savani university · surat, india
software engineering, IoT, blockchain
writing
context is the new compute
Apr 2026in march 2026 anthropic shipped a 1 million token window at standard pricing and inference costs collapsed 240x in 18 months. compute is cheap. context is expensive. here's the discipline that actually separates production agents from demos in 2026.
the 25% agent: designing for the reliability we actually have
Apr 2026a public benchmark ran a 6-task crm agent flow 10 times and hit 25% end-to-end success. meanwhile 57% of orgs claim agents in production. that gap is the most important engineering problem in 2026. here's how i design for the reliability we actually measure.
the code that outlasts you isn't the code you wrote
Apr 2026stewardship is the engineering virtue nobody budgets for. but the thing that actually outlasts you isn't the code, it's the constraints, the names, the data model, and the 'why' doc. and the best act of stewardship is usually deletion.
projects
regent
↗ai executive assistant platform
a production-grade SaaS platform that replaces human executive assistants with per-user ai agents. Go backend orchestrating agentic workflows: 13-stage email processing (categorize, summarize, RAG-backed drafting), multi-model LLM routing with circuit-breaker fallback, and goroutine supervision with backoff recovery. 107-component Next.js frontend with 27 routes, 5 Zustand stores, real-time Supabase WebSocket subscriptions, 60 SQL migrations with compile-time RLS, AES-256-GCM encryption with key rotation, 4-tier Stripe billing ($97 to $697/mo), and 16 scheduled cron jobs for nightly batch processing.
researchconnect
↗ai research collaboration discovery
ai-powered platform for UT Arlington's MAE department that matches researchers using semantic vector search. FastAPI backend converts natural language queries into 384-dim embeddings, searches faculty profiles via pgvector, queries OpenAlex/arXiv/NSF in parallel, and uses LLMs to explain match rationale, all streamed via SSE. Next.js 16 frontend with Supabase auth and Playwright E2E tests.
mailpilot
↗intelligent email automation
email automation platform with real-time streaming architecture. Go backend runs async job workers for email fetching, ai processing, and sending, with a cron scheduler for periodic IMAP sync and digest delivery. SSE broker pushes live pipeline status to the Next.js frontend. dual api pattern: server-side REST proxy for mutations, client-side SSE for real-time updates.
skills
programming languages
frameworks & libraries
databases
tools & software
cloud platforms
security
certifications