Bharat Kumar AdhikariSenior Full-Stack & AI EngineerSenior Full-Stack · Payments · Identity · Fraud · AI
11+ years building production platforms across payments, identity verification, fraud detection, e-commerce, and AI.
System map
02 — Production domains
Payments, identity, fraud, commerce and AI — with the products and skills that sit in each domain.
Payments
Money movement that keeps going when a vendor fails, changes state, or times out.
Identity / KYC
The checks that decide whether a person can move money — wired into the product, not beside it.
Fraud
Fraud signals on the request that matters — across five transaction types.
eCommerce
Shopify, WooCommerce and custom commerce flows — catalog, checkout and the APIs behind them.
Modernization
Keep the schema and the rules that already run the business. Replace one layer at a time.
AI infrastructure
Routing, circuit breakers, cost and RAG — the same reliability patterns used on payments.
Semantic video
Meaning-based scenes, Remotion render, and providers labelled when they are mock or missing.
Content & engagement
AI inside WordPress, YouTube mention search, and a Gemini chatbot that captures out-of-scope leads.
How the systems stay up
03Watch
Play the video, then download the free ebook if you want the full 7-step playbook.
04What I solve
Production payments, identity verification, fraud detection and compliance — with provider abstraction and failover when vendors fail or change state.
Modernize aging applications without throwing away the business logic and data that already work.
Introduce LLMs, RAG and automation where they improve a real workflow.
Build storefronts, APIs, integrations and customer-facing experiences designed around the commerce journey.
05Systems thinking
Modernization isn't one big move. It's a sequence of layers — each one reduces risk and unlocks the next. Select a layer to see what it solves.
Business rules and years of data live in an aging codebase that nobody wants to touch — so every change is slow and risky.
It already runs the business. Understanding it first is what prevents a rewrite from quietly dropping behaviour customers rely on.
06Featured work
Case studies you can inspect — each card opens a write-up built only from the repository and the public site.
AI Infrastructure
One gateway between your applications and OpenAI, Gemini and Anthropic — with routing, failover, cost tracking and grounded retrieval built in.
Read case study
Source (opens in a new tab)
Legacy Modernization
A Node.js and React rebuild of a CodeIgniter freelancing marketplace that keeps the existing MySQL database and business rules intact.
AI Product · Semantic Video
An audio/text → visual-story platform that understands meaning, plans scenes, matches assets and renders captioned MP4s with Remotion.
Read case study
Source (opens in a new tab)
AI Product · Media Intelligence
A full-stack platform that takes a YouTube video and an advertiser name and returns timestamped mentions, transcript context and a link that opens at that exact moment.
AI Product · WordPress
A WordPress plugin that scans content on publish, scores SEO health and suggests fixes with OpenAI — inside a monthly budget and with human approval.
Read case study
Source (opens in a new tab)
AI Product · Lead Generation
A Gemini-powered website chatbot that answers from site context and turns out-of-scope questions into captured leads with instant Telegram alerts.
AI Product · Browser Extension
A Chrome extension and mobile web app that summarizes webpages and YouTube videos with Gemini — right where the reader already is.
Read case study
Source (opens in a new tab)
eCommerce · Shopify
Shopify storefront engineering for a live consumer brand — customer-facing commerce work in production.
Read case study
Live (opens in a new tab)07Flagship case study · AxiomAI
One gateway between your applications and OpenAI, Gemini and Anthropic — with routing, failover, cost tracking and grounded retrieval built in.
Designed request path
01 / 06
Application. A client sends a chat request over HTTP. It never talks to a vendor SDK.
Applications that call LLM vendors directly inherit every vendor's outages, pricing and SDK quirks. Keys leak into services, costs are invisible until the invoice arrives, and there is no single place to enforce limits or ground answers in company knowledge.
Applications never talk to model vendors directly. Every request passes a linear, individually testable pipeline before a deterministic router chooses a provider adapter.
08Modernization story · Foxlancer
A Node.js and React rebuild of a CodeIgniter freelancing marketplace that keeps the existing MySQL database and business rules intact.
Strangler migration
schema stays · modules move
Existing MySQL. serv_* tables stay. Both stacks read and write the same data.
A working freelancing marketplace — jobs, bids, hiring, wallets, milestones, escrow and disputes — ran on an aging CodeIgniter codebase. A big-bang rewrite with a new database would put years of data and business rules at risk.
The legacy application is left untouched as the reference implementation. The new stack reads and writes the same MySQL tables, so both can be compared side by side while modules move across.
Gaps are listed in the public migration tracker so progress is not mistaken for completion.
09AI application engineering
Four moves: understand the material, attach it to systems you already run, automate one step, then operate it like production software.
Understand
Ground the model in text that belongs to the business, then retrieve only what the next step needs.
10Engineering capabilities
Not a logo wall. These are the languages, systems and domains that show up in the work above.
11How I think
The running system is the specification. Read the data and the rules before you replace a line.
Foxlancer keeps the legacy app as referenceA model earns its place only when it changes a specific workflow — support, content, leads, retrieval.
LeadGenie turns out-of-scope questions into leadsIf you cannot see hops, failures and cost, you do not have a product. You have a demo.
AxiomAI persists usage, cost and every provider hopSearch and traffic numbers come from the source system. Models do not invent them.
Vantage SEO reads Search Console, never generates itFake providers, limits and a human approval step are how a local project stays honest.
AxiomAI's 80+ pytest cases run on fake providers12How I work
Understand the system, map the seams, modernize behind a contract, add intelligence to one workflow, then operate it.
Read the existing system: the tables, the rules, and the behaviour customers already rely on.
Find the seams. Separate what must stay from what can move, module by module.
Replace one layer at a time behind a stable contract. Keep the database when it is the business.
Attach a model to one workflow, grounded in your data, with a human where judgement matters.
Limits, failover, cost and isolation come before calling it production.
13 — AI Transformation Blueprint
The 7-Step AI Transformation Blueprint
A practical guide for eCommerce teams looking to modernize legacy systems, introduce AI and improve workflows without unnecessary rewrites.
14Open source
Public repositories for the systems on this site. Read the case study when you want the decisions; clone the repo when you want the implementation.
One gateway between your applications and OpenAI, Gemini and Anthropic — with routing, failover, cost tracking and grounded retrieval built in.
A local-first control plane: routing, failover, cost tracking and cited retrieval in one place.
An audio/text → visual-story platform that understands meaning, plans scenes, matches assets and renders captioned MP4s with Remotion.
Audio or text becomes a meaning-planned visual story — Remotion renders it, unconfigured providers stay labelled.
A full-stack platform that takes a YouTube video and an advertiser name and returns timestamped mentions, transcript context and a link that opens at that exact moment.
Timestamped advertiser mentions in long-form YouTube — captions first, Whisper only if needed.
A WordPress plugin that scans content on publish, scores SEO health and suggests fixes with OpenAI — inside a monthly budget and with human approval.
AI-assisted SEO inside WordPress — with a monthly budget and a human approval step.
A Chrome extension and mobile web app that summarizes webpages and YouTube videos with Gemini — right where the reader already is.
Summarization where the reader already is: a Chrome sidebar and a mobile web app.
A Gemini-powered website chatbot that answers from site context and turns out-of-scope questions into captured leads with instant Telegram alerts.
Website-grounded answers; out-of-scope questions become captured leads.
A Node.js and React rebuild of a CodeIgniter freelancing marketplace that keeps the existing MySQL database and business rules intact.
A strangler migration that keeps the existing MySQL schema and business rules.
A Flask application on PostgreSQL, deployed to Render with render.yaml and runnable locally with Docker Compose.
A small Flask service whose deployment is described in code.
15About
11+ years building production platforms across payments, identity verification, fraud detection, e-commerce, and AI — not a single vertical.
I am a full-stack engineer who ships production software, then applies the same reliability patterns to AI. The work lives where money, identity, fraud, storefronts and models meet real business logic.
My work spans the Socure identity verification lifecycle — DocV, ReSelfie, and deposit-threshold gating — SEON fraud detection across five transaction types, and integrations with six payment providers: Stripe, Breeze, Checkbook/ACH, AptPay, Fiserv DDP, and Prizeout.
Production depth
Identity
DocV, ReSelfie, and deposit-threshold gating — the checks that decide whether a person can move money.
Fraud
Fraud signals wired into the transaction path, not a dashboard sitting beside it.
Payments
Stripe, Breeze, Checkbook/ACH, AptPay, Fiserv DDP, and Prizeout — routed so a vendor outage does not freeze the product.
AI + Production Engineering
I'm applying the same engineering patterns I use in payments and compliance — routing, abstraction, failover, reliability, and observability — to production AI systems.
Core stack
What I bring
Not an AI researcher. Not a prompt engineer.
What differentiates me is the combination of production depth in payments, KYC/identity verification, fraud, and transaction systems with hands-on experience building and deploying AI applications and LLM infrastructure.
I'm an engineer who builds systems where AI meets real-world business logic, integrations, compliance, and reliability.
Build intelligent systems. Engineer for reliability. Design for scale. Ship for real-world impact.
16Contact
Let's understand the problem first, then choose the technology.
Email bk.digitalservice@gmail.comPhone +91 98309 82129