Full-Stack & AI Product Engineer — Remote, India

Subham
Dixit.

Four plus years turning ambiguous requirements into production TypeScript, React, Next.js & Node systems. Right now, building and scoring AI models under the hood of ReelAutopsy — a live SaaS I designed, built, and launched solo, from schema to sign-up, at a scope usually split across a 3–4 person team.

View ReelAutopsy → Get in touch
4+ yrs in production
6 engineering roles
1 solo SaaS, live
4★ CodeChef, top 2%
REELAUTOPSY // 8-AXIS SCRIPT SCORING MODEL — DRAG TO ROTATE
01 — FLAGSHIP PRODUCT

ReelAutopsy

An AI-powered Instagram Reels script analyzer, built end-to-end as a solo founder — architecture, backend, billing, and go-to-market. Creators paste a script and get back a structured diagnostic: what's working, what's costing them the scroll, and why.

Every score is generated through a schema-validated pipeline rather than free-text AI output, so results stay consistent and structured across thousands of scripts — the same rigor Subham brought to evaluating LLM response pairs as an AI Trainer. Grew to 50 signups through 100% organic acquisition (Reddit, zero ad spend), with zero-downtime deploys and no reported outages since launch.

~4.5s
Avg. scoring response
95%+
Schema validation rate
50
Organic signups
Solo
Built & shipped by
Production Stack
Next.js App Router→ Gemini API→ Zod validation
Prisma→ Neon Postgres→ Upstash Redis
Google OAuth→ Resend→ Razorpay billing
HMAC webhooks→ Vercel CI/CD
Scoring dimensions — sample of 8
01
Hook Strength
02
Scroll Risk
03
CTA Effectiveness
04
Pacing
05
Structure
06
Clarity
07
Retention Signal
08
Audience Fit
02 — CAREER TIMELINE

Experience

JAN 2026 — PRESENT
Founder & Full-Stack Developer
ReelAutopsy · Remote
  • Architected an AI scoring pipeline (Gemini API + Zod schema validation) rating scripts across 8 dimensions — hook strength, scroll risk, CTA effectiveness, pacing — with ~4.5s average response time and 95%+ successful schema validation across production API calls.
  • Designed and shipped a production-grade SaaS solo — architecture, authentication, billing, and deployment — a scope typically split across a 3–4 person team, using Next.js App Router, Prisma ORM, PostgreSQL (Neon), Razorpay payment integration, Upstash Redis async job queues, Resend, Google OAuth, and HMAC-verified webhooks.
  • Grew ReelAutopsy to 50 signups through 100% organic acquisition — Reddit community engagement, zero ad spend — validating early product-market signal pre-funding.
  • Owns the complete product lifecycle from architecture to deployment (Vercel CI/CD), with zero-downtime deploys and no reported outages since launch.
Stack: Next.js, React, TypeScript, Gemini API, Zod, Prisma, Neon Postgres, Razorpay, Upstash Redis, Resend, Google OAuth, Vercel
MAY 2025 — JAN 2026
AI Trainer
Soul AI · Remote
  • Evaluated pairs of LLM responses on correctness, clarity, and helpfulness, authoring structured rationale across 100+ evaluations to support model improvement workflows.
  • Identified failure modes and edge cases in model outputs, proposing prompt-engineering fixes that reduced error rates in follow-up evaluation rounds.
  • This evaluation discipline directly shaped the design of ReelAutopsy's own scoring pipeline.
AUG 2024 — APR 2025
Software Developer
Virtuelly Inc. · Remote
  • Built an AI Photobooth solo on Node.js (AWS Lambda) and the Replicate API, architecting an async serverless pipeline for concurrent styled-avatar generation; boosted post-event engagement by 30%.
  • Delivered a real-time multiplayer game from scratch in React and Firebase, deployed across 50+ corporate events supporting 200+ concurrent players.
  • Reduced enterprise event page load time from ~10s to ~1.4s (86% improvement) by replacing pre-rendered pricing combinations with input-driven logic.
  • Designed secure APIs for session moderation, scoring, and event analytics.
Stack: Node.js, AWS Lambda, Replicate API, React, Firebase
JUN 2024 — JUL 2024
AI Trainer (Freelance)
Remotasks
  • Assessed and ranked AI-generated response pairs against quality criteria, flagging errors and edge-case failures to support iterative model improvement.
JAN 2024 — MAY 2024
Software Engineer
Procurenet · Remote
  • Extended an LLM content-generation pipeline with Hindi-language support via a new prompt producing structured Hindi articles alongside the existing English flow.
  • Enhanced content discoverability by integrating Zilliz Vector DB for similarity-based clustering and semantic retrieval across the article corpus.
  • Built a keyword-based duplicate-content check that runs at publish time.
APR 2022 — DEC 2023
Software Engineer
Virtuelly Inc. · Remote
  • Joined as a junior engineer on a 7-person team building enterprise virtual event platforms (Node.js, React, AWS) serving thousands of concurrent users.
  • Owned scoring logic and built REST APIs with Socket.io powering large-scale virtual events, handling thousands of users with low latency.
03 — STACK

Technical Skills

Frontend
ReactNext.js (App Router) TypeScriptTailwind HTML / CSS
Backend & APIs
Node.jsREST API Design Auth (OAuth)Prisma Python (Flask)Serverless AWS LambdaAsync Job Queues Webhooks3rd-Party API Integrations
Data & Storage
PostgreSQL (Neon)Redis (Upstash) MongoDBMySQL Zilliz Vector DB
Cloud & DevOps
AWS (Lambda, S3, ECR, CodeBuild)Vercel DockerCI/CDGit/GitHub
AI & LLM Products
Gemini APIOpenAI LangChainReplicate ZodSocket.io Firebase
AI Dev Tools
Claude CodeCursor PostmanSlack Vercel ObservabilityUptimeRobot
04 — RECORD

Achievements & Education

Competitive Programming
Top 2% · Rank 439, CodeChef January Long Challenge 2021. 4★ Coder, rating 1806.
Recruitables 2021
Finalist among 100,000+ applicants — top 6%, Dare2Compete Recruitables.
Education
B.Tech, Computer Science & IT — IPS Academy (IES), Indore. 2018–2022, CGPA 8.5/10.