← Back to selected work

Aadhar Gupta

Full-stack & Applied AI Engineer

Noida, Delhi NCR, Indiaaadhar02gupta@gmail.comLinkedIn / aadhar02guptaGitHub / Aadhar-Gupta

Product engineer working across enterprise search at Algolia, full-stack client products and applied AI. Builds interfaces, services, data pipelines and agent workflows with deliberate permissions, evaluation and delivery controls. Uses AI coding agents daily while owning architecture, review and outcomes.

Experience

Data Engineer — Algolia

Contract via Ayasya Digital Solutions · Nov 2024–present · Remote

  • Authored 98 storefront search/commerce integrations across branches, with 69 on main, normalizing search, browse, cart and purchase events for BigQuery (Oct 2026 repository snapshot).
  • Improved shared connector routing, anonymous-device handling and event deduplication; shipped changes through a CDN release train.
  • Standardized integration patterns, bringing delivery from roughly a week to two days (personal delivery estimate).

Software Engineer — Ayasya Digital Solutions

Official designation: Associate Developer (L1) · Full-time Mar 2024–present
Internship Sep 2023–Mar 2024 · Training Mar–Sep 2023

  • Contributed to a multi-tenant AI SaaS platform spanning 39 services and a shared web/iOS/Android app: agent workflows, four MCP servers, human approvals, structured email planning, a canvas editor and shared asset flows.
  • Hardened CI/CD with smoke-triggered rollback, delivery guards and test sharding. Improved a marketing site’s measured mobile PageSpeed score from 64 to 86.
  • Main web/API engineer for a healthcare booking product: payment deduplication, booking concurrency, consent and deletion controls; cut Android package size from 110.6 to 45.6 MiB.
  • Directed an AI-assisted design-system-to-Jahia migration workflow: 50 marketing pages on one site and 63 pages on a second, with Storybook and accessibility/parity checks.

Selected builds

hOS — Retrieval & agent platform

Solo-built prototype · Python, FastAPI, PostgreSQL/pgvector, LiteLLM, MCP

Built hybrid retrieval with tenant/access/date filters, citation access checks, encrypted credentials and Postgres-backed jobs. 757 test functions in 105 files. Reduced assistant turns from 40–46 s to 8–13 s by correcting embedding precision and memory pressure. Began as a company POC; continued as a personal assistant.

Relationship intelligence platform

Solo-built, pre-launch · Fastify, Next.js, Postgres RLS, Drizzle, Vercel AI SDK

Built tenant-scoped deterministic retrieval, a structured planner, approval-gated outreach, usage metering and eight evaluation suites. Achieved 11.4× fewer tokens on a typical synthetic workload; this is a benchmark, not a production result.

Technical skills

Interfaces
TypeScript, React, Next.js, React Native / Expo, Storybook, accessible UI
Services & data
Node.js, Express, Fastify, Python / FastAPI, Java / Spring, PostgreSQL, pgvector, RLS, BigQuery, Redis
Applied AI
Claude, MCP, hybrid RAG, structured generation, evaluations, budgets, LiteLLM, human approval workflows
Delivery
Google Cloud Run / Cloud SQL, Cloudflare Workers, Docker, Terraform, GitHub Actions, Playwright, Jest, pytest, OpenTelemetry

Education & certifications

BCA — Chaudhary Charan Singh University · 2022–2025 · First Division

Algolia Certified Developer · Algolia Search Engineering Basics · Anthropic Claude 101 · Introduction to MCP · Introduction to Agent Skills