Hello! I'm Ashlesh.
Building systems
that scale.
I build and scale production-grade backend systems — event-driven, serverless, and distributed. I've led teams, owned core architecture, and delivered end-to-end as a technical lead.
3
Years building backend systems
6
Client projects delivered end-to-end
Tech Stack
Skills & Tools
Backend & Architecture
AI & LLM
Cloud & DevOps
Auth & Security
Databases
Frontend & Mobile
Work History
Experience
2024 — Present
Consulting
01
Domi Inc.
Founding Backend EngineerMailer platform for US mailers to send physical mails with legal compliance and auditing. Empowers mailers to run targeted campaigns based on user interest.
- —Led architecture for 4+ applications and handled DevOps across 3 environments on AWS
- —Built RAG features, LLM integrations, real-time sockets & chat
- —Built multi-step onboarding, deletion, archive & transfer workflows with OTP, token verification, and document verification
- —Stripe integration: subscription management and connected accounts for platform payments
- —Managed CD pipelines (GitHub Actions), ECS + Docker for backend, frontend, and mobile
- —Built 3rd-party webhooks and internal webhooks using shared database architecture
- —Implemented RBAC and user management via Supabase; Redis for caching and process verification
- —AI automation that reduced manual efforts by 80%
- —Clean code MVC architecture; Swagger docs and Jest unit tests; Sentry + CloudWatch for logging
2024 — 2025
Consulting
02
Skyclad Ventures
Technical LeadSpent one year as a partner in a consulting firm, leading two client projects end-to-end as a technical lead, driving successful delivery for maritime and compliance industries.
- —Led a team of 4 developers, providing technical direction and ownership of core deliverables
- —Smart PMS dashboard: serverless architecture on Supabase with Edge functions, DB functions, RLS, and Postgres — covering ship, device, survey, and user management with RBAC for 6+ roles
- —Document Intelligence Platform: two-layered OCR + LLM pipeline for 200–300 page maritime documents (manuals, invoices, RFQs) achieving superior accuracy vs Omni AI
- —Evolved architecture from HTTP → polling (BullMQ + Redis) → webhook-based with exponential backoff, HMAC security, and response logging
- —Built API-as-a-service with usage tracking, Swagger docs, Redocly user docs, and Metabase monitoring dashboards
- —RAG solution for querying PDFs, DOCX, and images using Mongoose vector search with cosine similarity + keyword search
- —NL-to-SQL query interface for enterprise databases with 600+ interconnected tables
- —Deployed via EC2 + Nginx, S3, AWS Amplify; CI/CD with GitHub Actions; Stripe subscription integration
2026
Consulting
03
HireRight
Lead AI consultantConsulted to design and build an AI-powered interview software platform targeting non-technical job roles — reducing pre-screening costs by 40% and fully eliminating manual pre-interview efforts.
- —Led a team of 3 developers end-to-end through design, build, and delivery
- —Built real-time AI interview system with WebRTC streaming and egress
- —Implemented AI agent orchestration and automated evaluation pipeline
- —Pre-screening costs reduced by 40%; human effort for pre-interviews fully eliminated
2025
Consulting
04
Veriso
AI ConsultantConsulted an EU-focused audit firm to automate their SMETA compliance review process — removing the need for teams to manually review hundreds of pages of audit documents.
- —Designed and delivered an automated pipeline for extracting and analysing SMETA audit PDFs
- —Identified compliance issues and non-conformances programmatically from audit reports
- —Generated and validated Corrective Action Plan Reports (CAPR) in Excel
- —Assessed and classified severity levels against NC guidelines, local laws, and regulations
- —Outcome: massive time savings and freed the client's team for higher-value strategic work
2025 — Present
Consulting
05
Medexa
Backend EngineerEnd-to-end backend for a multi-tenant hospital management SaaS — 13 independently deployed serverless stacks covering the full clinical surface: lab orders, surgery workflows, OPD/IPD, and radiology.
- —Serverless microservices: 13 independently deployed stacks with ~594 Lambdas, per-function esbuild bundling, and a full local dev stack
- —Terraform IaC: provisioned RDS, RDS Proxy, S3, IAM, Secrets Manager, and Security Groups across dev/stage/prod with strict variable validation
- —RDS + RDS Proxy: mutual TLS enforcement, two-layer security groups (Lambdas never hit RDS directly), Secrets Manager credential rotation, 30-min idle timeout
- —Dual API Gateway setup: main gateway (admin/clinical ops) + clinical gateway (lab/surgery/radiology) with separate rate limiting (5k burst / 2k rps), CORS policies, and CloudWatch logging
- —658+ REST APIs covering the full hospital surface: 8-step lab order lifecycle, multi-phase surgery state machine, OPD queues, IPD workflows — backed by 430 Zod validation schemas
- —Cognito JWT auth at gateway layer + Casbin RBAC at app layer (runtime policy changes, no redeployment needed)
- —Multi-tenant SaaS architecture
Selected Work
Projects
SaaS · Live
Watomate
WhatsApp automation for Shopify
Powerful automation platform for Shopify store owners — supercharging pre-sales, post-sales, and loyalty funnels via WhatsApp. Inspired by Postscript, built for WhatsApp.
- —Customer segmentation via CSV or manual input
- —AI-powered dynamic message rephrasing per audience style
- —Shopify integration in 2 clicks
- —Event-based automation triggers
- —AI-generated relevant images for campaigns
- —Privacy-first: all data wiped on sign out
GEN AI · Live
Clipmailo
AI newsletter generator
Generate and publish thoughtful AI-generated newsletters in under 2 minutes. Fine-tuned on Mistral with web scraping for real-time content injection.
- —Mistral fine-tuned for newsletter generation
- —Data injection via web scraping for latest content
- —10+ beautiful templates
- —Login with YouTube — generate from YouTube content
- —Audience & subscription list management
- —Google Auth implemented from scratch
AI · Personal
findmydress
Visual search via WhatsApp & Instagram
WhatsApp and Instagram bot that searches the web for fashion items. Share a reel, image, or text with @search — the bot finds matching products across the web.
- —CLIP models for unified image + text vector space
- —Extended CLIP to embed metadata (price, description, rating)
- —Dual embedding collections with reranking and deduplication
- —Evaluation scripts for retrieval quality and LLM responses
- —Scraped Myntra and Meesho for dataset
Backend · Personal
Notification System
Realtime high-volume notifications
Helps large ERP-like software handle notifications at bulk — scales to millions of notifications with millisecond delivery latency.
- —WebSocket-based real-time delivery
- —Message queuing + microservices for high-volume processing
- —JWT authentication and authorization
- —Dockerized services for easy deployment
- —Comprehensive Swagger documentation
Backend · Personal
Video Streaming
Adaptive bitrate streaming on AWS
Streams high-volume, high-scale videos with adaptive bitrate and an encoder that auto-scales with user traffic. 99.99% uptime with a retry queue mechanism.
- —MPEG-DASH adaptive bitrate streaming via FFmpeg
- —Decoupled architecture: SNS, SQS, and Lambda
- —EC2 Auto Scaling Group with CloudWatch step scaling
- —Low latency delivery via CloudFront CDN
- —OAuth 2.0 Google social login
- —Secure secrets via Parameter Store and IAM roles
Mobile · Personal
NutriFit
Android calorie & nutrition tracker
Android app that helps users trace daily calories, micro and macronutrients, and set fitness goals — similar to HealthifyMe. Built natively in Java.
- —Daily calorie and macro tracking
- —Goal setting and progress monitoring
- —Native Android (Java) — available on GitHub
Co-Founder
Eazy Cheez
Co-founded a hyper-local food & delivery platform — a “Zomato for tier 2 & 3 cities” — built on a non-commission model to help restaurants, grocery stores, and vendors deliver within a 2 km radius.
The 2 km constraint and freelanced delivery model kept costs low for price-sensitive markets across Indian tier 3 cities. A custom algorithm was built for efficient order-to-delivery allocation.
12
Vendors onboarded across ~15 towns near Mangalore
70+
Users onboarded on the platform
₹5L
Government grant received
3
Interns and employees led from scratch