Akshat_Soni
SYSTEM_ARCHITECT0→1_BUILDERDISTRIBUTED_SYSTEMS

SYS_LOG: 0.0% // SECURE_CONN

DELIVERING

FULL_STACK_SYSTEMS
0→1 PRODUCTION READY

$I ARCHITECT FULL-STACK SYSTEMS, DYNAMIC FRONTENDS, & HIGH-VELOCITY BACKENDS FOR 0→1 VENTURES

GITHUB_PROFILE
START_COLLAB
Loc: REMOTE_GLOBAL // UTC+5:30
Section_01

Engineering History.

Jan 2025 — Present

Founding Engineer

@Lexomat

  • Environment & Domain: AI/LLM Infrastructure, RAG Pipelines, Cloud-Native Systems, Distributed Task Orchestration, Vector Databases.

  • Architected and shipped a production AI legal-tech MVP from 0→1 in just 2 months, owning the entire stack from infrastructure to frontend.

  • Engineered a high-throughput RAG pipeline (FastAPI, pgvector) querying against a highly optimized vector database scaled to over 4.5+ million legal documents and embeddings.

  • Engineered a cloud-native, asynchronous orchestration layer using TaskIQ and Redis, decoupling high-latency LLM workflows to guarantee zero data loss and system stability during client-side disconnects.

  • Architected a continuous evaluation CI/CD pipeline featuring automated E2E testing for non-deterministic LLM workflows, establishing strict guardrails against prompt regressions and retrieval failures.

  • Integrated Stripe billing and automated invoice generation system using headless browser rendering and backend automation pipelines.

  • Orchestrate the full system lifecycle—from architecture design to deployment and monitoring—ensuring high production reliability and rapid bug resolution.

July 2023 — Dec 2024

Software Engineer

@eSaral Ventures

  • Environment & Domain: High-Traffic Web Systems, React SPA Architecture, Backend Automation, AWS Lambda Serverless Computing.

  • Built fault-tolerant quiz frontend for 10,000+ DAU: leveraged IndexedDB and BroadcastChannel API to ensure zero session loss across multi-tab browser sessions.

  • Executed a solo 15-day complete frontend migration from a legacy monolith to a modern React SPA: reduced page load times from 3s to under 1s, improved Core Web Vitals, and unblocked a 200+ member engineering organization to accelerate rapid feature iteration.

  • Automated document workflows via AWS Lambda and Python: eliminated manual processes and cut operational overhead by 80%.

  • Shipped critical features for a cross-platform React Native mobile application, improving the learning experience for thousands of active students.

Section_02

Impactful Builds.

Architecture_History // Live_Pulse

Project_01// FEATURED

Longshot — Replay-Then-Tail SSE

Read_The_Deep_Dive
Sanitized extraction of an architecture I built at Lexomat for long-running LLM pipelines: 5+ minute AI jobs that survive client disconnects via replay-then-tail SSE, dedup retries via an idempotency lock, and emit strictly-ordered progress through a per-session async drainer.
Demonstrates the failure modes that matter in production — worker crash, network drops, mid-task cancel, concurrent emits, race conditions in pub/sub-vs-list ordering — with the fixes verified by a passing test suite.
Engineering deep-dive on dev.to — how to design replay-then-tail SSE, why three common fixes for the seq-ordering race fall short, and the asyncio drainer pattern that solves it without distributed locks.
FastAPIRedis StreamsTaskIQPydanticSSE
Read_The_Deep_Dive
Build_Ref_LON
Project_02

AI SaaS Platform

Built a full-stack AI application (Next.js, Prisma, Stripe) to experiment with early LLM capabilities, implementing foundational backend authentication and basic subscription workflows.
Integrated LLM APIs for core functionality and AI-driven features.
Implemented rigorous backend subscription enforcement and precise usage control.
Next.jsTypeScriptPrismaStripeLLMs
View_Source_Code
Build_Ref_AI
Project_03

Python IDE Platform

Developed a fully-functional browser-based Python IDE featuring a secure backend execution environment.
Engineered authenticated code storage for users leveraging Django and AWS infrastructure.
DjangoReactAWSPython
View_Source_Code
Build_Ref_PYT
Project_04

Vector Image Search

Developed a high-performance image search application leveraging vector embeddings for semantic, similarity-based retrieval.
Integrated Supabase as the vector database backend to efficiently query large dimensional spaces in near real-time.
Next.jsSupabaseTypeScriptTailwind CSS
View_Source_Code
Build_Ref_VEC
Project_05

Dynamic Form Builder

Engineered a drag-and-drop form building interface using dnd-kit, allowing intuitive construction of customizable component structures.
Architected the database schema using Prisma and Supabase for real-time synchronization, with Clerk handling secure, multi-factor authentication.
Next.jsPrismaSupabaseDND-KitClerk
View_Source_Code
Build_Ref_DYN

Explore Repository

Constant iteration across distributed systems and AI experiments.

@akshatsoni26
FETCHING_REMOTE_DATA...