// based in Bengaluru, India

Abhishek
Srivastava

Full-Stack Engineer Applied AI

I build multi-tenant backends and ship generative AI into production — RAG pipelines, image generation, and segmentation models running reliably on Azure. 2+ years across products for clients in Singapore, the UK, and an Indian PSU.

View my work View résumé Get in touch
GitHub LinkedIn X / Twitter
01

About

I'm a Software Engineer at Misfits Communications, working across the stack for international product clients and Indian public-sector platforms. My center of gravity is backend architecture and applied AI — designing multi-tenant systems, REST/GraphQL APIs, and getting models like Gemini, OpenAI, and SAM to run reliably in production.

I'm currently completing an MCA specializing in AI/ML alongside full-time work, and I care about the unglamorous parts: fault isolation, latency budgets, and clean service boundaries that let teams ship independently.

  • 2+years shipping production systems
  • 10microservice AI backend on Azure
  • 3countries' clients served
02

Experience

Mar 2025 — Present

Software Engineer · Misfits Communications

Full-stack and generative-AI engineering

  • Helped split a monolithic workflow into an orchestrator + specialized microservices (Query Intent, Workflow, RAG Retrieval, Image Generation…), enabling independent deployment.
  • Built a FastAPI service unifying virtual try-on, floor visualizer, and wall-paint on one port with fault-isolated sub-app mounts.
  • Shipped SAM 3 + MobileSAM + Segformer masking with tap-to-refine UX; cut try-on first-view latency ~29s → ~11s.
  • Rebuilt RVNL's legacy site into MERN with an 81-table bilingual (EN/HI) schema, JWT + RBAC, audit logs, and an approvals workflow.
May 2024 — Feb 2025

Python Developer · MP Research Work

Backend engineering on client-facing REST APIs.

  • Designed and shipped CRUD REST APIs in Python (FastAPI, Django, Flask) with clean resource routing, consistent response envelopes, and proper HTTP status semantics.
  • Modeled data with SQLAlchemy/ORM and enforced request/response contracts using Pydantic schemas for validation, serialization, and auto-generated OpenAPI docs.
  • Implemented pagination, filtering, and sorting on list endpoints, plus token-based auth and role checks to guard create/update/delete operations.
  • Built third-party integrations and leaned on FastAPI's async endpoints and dependency injection for testable, non-blocking I/O.
  • Optimized PostgreSQL query performance and wrote PyTest unit/integration suites to reduce regressions across releases.
03

Featured Projects

Production · Company · Solo-built services GenAI Platform

Agrim AI

A Shopify-integrated generative-AI e-commerce platform — a 10-microservice AI backend behind a WAF + Application Gateway on a segmented Azure VNet. Within it, I independently designed, built, and own two production microservices end-to-end:

Image Service Virtual Try-On · Floor · Wall Paint

A unified FastAPI service that consolidates three visual product-visualization capabilities into one containerized deployment — turning a person + garment, a room + tile, or a wall + paint into photorealistic previews before purchase.

  • Consolidated three capabilities — Virtual Try-On, Floor Visualizer, and Wall Paint — into a single deployment, each isolated in its own module with graceful degradation: if one fails to initialize, the others keep serving, and /health reports per-capability readiness.
  • Generated three photorealistic views (front / side / back) per try-on via Gemini, with an "Identity Guard" prompt that locks the customer's face, hair, and body and treats the garment purely as source — eliminating the common identity-swap failure.
  • Built a five-layer safety gate (person-presence, minor detection, nudity, garment validity, lighting) with fail-closed safety checks and fail-open quality checks, plus a post-generation pass on the model's own output.
  • Added confidence-based non-wearable filtering so décor/objects can't hallucinate into garments, with batched VLM prechecks (8 concurrent, verdict cached per product) so storefront grids never hammer the model.
  • Engineered two-tier wall masking — MobileSAM point-prompted precision gated by a Segformer wall-class mask, with auto wall-detection over a probe grid and IOU dedup; floors use Segformer + a perspective-aware warp to re-tile a room.
  • Cut try-on first-view latency ~29s → ~11s via progressive delivery, and used downscale-then-composite to keep Gemini floor renders around ~15-20s.

Integration Service API Gateway · Webhook Router

A configurable API-to-API gateway that decouples the platform from third-party services — clients register an integration once as a database "recipe", with no code changes or redeploys to onboard a new external API.

  • Designed a recipe-driven gateway where each integration (endpoint, method, request template, expected response keys) lives as a MySQL row — zero-code onboarding of new third-party APIs per client.
  • Wrote a {{variable}} template engine supporting nested dotted paths and deep recursive leaf lookup, so one recipe adapts to varied upstream payload shapes without duplication.
  • Verified responses by deep-searching required keys and confirming each is actually filled (not null/empty), with automatic configurable retries and full failure-state capture (status, missing keys, upstream body) for audit.
  • Exposed dual execution modes — direct-by-ID and semantic lookup-by-client + bot — with a request_id threaded through the platform's central logging pipeline.
  • Tuned the connection pool for production (pre-ping, hourly recycle, strict 10-connection cap) with explicit pool-exhaustion 503 handling, shipping fire-and-forget structured execution logs to a shared logging service.
  • FastAPI
  • Python
  • Gemini
  • OpenAI
  • SAM 3
  • MobileSAM
  • Segformer
  • PyTorch
  • MySQL
  • SQLModel
  • Azure
Personal · Full-Stack SaaS Live ↗

Procure-AI · smartprocure.in

An AI-native procurement platform that automates the entire RFP-to-vendor-selection flow — extracting structured data from vendor proposals, comparing quotes with compliance-aware scoring, and tracking real-time cost savings across a tiered SaaS subscription model.

  • Dual-model AI pipeline: Gemini 2.0 Flash for fast structured extraction from vendor PDFs, Gemini 2.5 Pro for deep comparative analysis — with plan-based prompts scaling features per tier.
  • Indian GST compliance: GSTIN verification, real-time filing checks, and effective cost-of-ownership that accounts for 18% Input Tax Credit losses.
  • Multimodal PDF processing: vision-based line-item extraction with automatic math sanity-checks that flag totals-vs-line-items discrepancies, even on non-English documents.
  • RAG history engine: completed RFPs indexed in ChromaDB so teams can query past procurement in natural language, grounded in org-specific data.
  • Tiered SaaS with prompt-level gating: Free / Pioneer Pro / Enterprise plans inject plan-specific instructions straight into the AI prompts — scaling from basic price/timeline comparison up to full tax-risk analysis.
  • Turborepo architecture: a React + Vite frontend, a Node/Express API, and a Python/FastAPI AI service in one monorepo — containerized with Docker and deployed to Google Cloud Run via Cloud Build.
  • React
  • TypeScript
  • FastAPI
  • Node.js
  • Gemini
  • ChromaDB
  • PostgreSQL
  • Docker
  • GCP
Personal · Solo build Live ↗

wellfitHR · wellfit.fun

A lightweight, multi-tenant HR platform for small teams — leave requests, payslips, employee directory, and holidays, minus the enterprise bloat. Engineered for minimal running cost: a single VPS + Postgres for the API, with the frontend on Vercel's free tier.

  • Clean API/UI split: a reusable Go (Chi) REST API decoupled from a Next.js 15 / React 19 frontend, ready for a future mobile client.
  • Multi-tenant from day one: every table carries an org_id, so onboarding a new company needs zero code changes.
  • 5-tier RBAC hierarchy (Super Admin → IT Admin → HR Admin → Manager → Employee) with checks that stop users creating roles above their own.
  • Zero-friction ops: embedded SQL migrations auto-run on startup, and payslips render to branded PDFs entirely client-side — keeping the API dependency-free.
  • Cost-engineered to run near-free: the Go API + Postgres fit a single always-free VPS behind Caddy (automatic Let's Encrypt TLS), with the frontend on Vercel's free tier.
  • Mock-mode frontend: a complete in-memory mock API implements the same TypeScript ApiClient interface as the real one — so every screen is buildable without a backend, and the build fails if real endpoints drift from the mock.
  • Go
  • Chi
  • Next.js 15
  • React 19
  • PostgreSQL
  • Docker
  • Caddy
04

Tech Stack

Languages

  • Python
  • TypeScript
  • JavaScript
  • SQL

Backend

  • FastAPI
  • Node.js
  • Express
  • Django
  • GraphQL

Frontend

  • React
  • Next.js
  • Vite
  • Tailwind CSS

Data & AI

  • PostgreSQL
  • MongoDB
  • Redis
  • Chroma
  • OpenAI
  • Gemini
  • HuggingFace

DevOps & Cloud

  • Docker
  • Azure
  • GitHub Actions
  • Git

// let's build something

Have a project or a role in mind?

I'm open to conversations about backend architecture and applied-AI work. The fastest way to reach me is email.

abhishek0wb@gmail.com
GitHub LinkedIn X / Twitter