BruceLin Kithion

Consultant · Solution architect · Builder

Ideas into working products, fast.

Portrait of BruceLin Kithion

I'm BruceLin Kithion. I've spent over 23 years in enterprise technology, inside global firms and startups. I bring creative thinking and hands-on engineering to the same table, so an idea becomes something you can use, quickly. My focus today is agentic AI.

Start a conversation See the builds

Approach

AI that reasons, validates, and acts.

Autonomy with governance. Every AI decision should be contextual, explainable, and tied to a business purpose. The model proposes. The state machine decides what happens next.

One request, traced State machine
  1. S1INTENTParse the request and open a session
  2. S2RETRIEVEGraph query first, then keyword, then vector
  3. S3REASONThe model proposes an answer
  4. S4VALIDATECheck it against rules and graph constraints
  5. ↺ fails a check: back to S3, or hand off to a person
  6. S5ACTCall the tool, post the entry
  7. S6EXPLAINSay what was done and why
Deterministic Model reasoning

Model

Reason

LLMs and knowledge graphs work as a pair. The graph holds what is true about the domain. The model works out what the person is asking for.

Deterministic

Validate

A state machine controls the flow. Rules live in tables with effective dates and citations, so every outcome traces back to its source.

Deterministic

Act

Agents call real tools: ERP postings, payment runs, search. Approvals and segregation of duties stay in place where the stakes call for them.

Modern enterprises need AI-native architecture, where deterministic systems and generative reasoning work side by side. I design the hybrid systems that think, validate, and scale.

Builds

I still design and ship.

I think like a product person and build like an engineer. These are the builds I'd show you first.

Advisor · Private beta

Anvikara

Your business, connected. Your next step, clear.

Anvikara brings purchasing, sales, inventory, finance, and compliance into one connected way of working for Indian businesses. It shows what happened, what needs attention, and who acts next. I advise the team on product and architecture.

Modules
Buy · Sell · Move · Reconcile · Account · Comply
Covers
Purchasing and payables, sales and collections, inventory, reconciliation, accounting and close, tax and compliance
AI
Explains, suggests, and prepares drafts from the business's own records. Approval stays with people.
Compliance
GST and TDS review, built for India
Built for
Business owners, finance teams, and CA practices
Enterprise build · In production

AI product configurator

A conversational configurator for complex, technical products.

Customers describe what they need and the system works toward a valid configuration. It runs in two modes: a guided flow for people new to the catalogue, and an expert flow for specialists who already know what they want.

Stack
FastAPI, Neo4j, Redis, PostgreSQL, OpenAI
Control
S1 to S7 state machine with adaptive flow
Search
Cypher → Lucene → Vector → LLM fallback
Hardening
Connection pooling, query safety, and per-user session isolation, tuned under production load
Consulting · Active

Trading platform architecture

Architecture advisory for a brokerage-grade trading platform. The details stay confidential.

Venture · Architected

Signal AI

An agentic competitive intelligence system, designed and documented as a co-venture.

From the lab

When an idea interests me, I build it. Most of these went from a question to a working prototype in a short time.

  • ArchiMind AI ↗A suite for software architects that runs different types of architecture assessments with AI. It can take in streaming telemetry and observability data and describe the architecture as it currently stands.Live demo
  • Nurtura AI ↗Supports children's mental health with early assessments and guidance for parents and teachers, covering attention, learning difficulties, and screen habits.Live demo
  • MyPsych Companion AI ↗A mental health marketplace that uses AI to match people with the right therapist, counsellor, or wellness professional.Live demo
  • MedCompanion AI ↗A radiology assistant prototype that uses computer vision to flag possible anomalies in X-rays and scans for a radiologist to review.Live demo
  • ThunaiA mental health companion with multiple personas and live AI. React.
  • LeadRadarA lead-generation pipeline for local businesses. React, deployed on Vercel.
  • Smart Code MigratorAutomates refactoring and modernization of legacy code.Private LLM
  • Smart Document ChatKnowledge retrieval across unstructured documents.Private LLM
  • Natural Language QueryingTurns plain-language questions into SQL.Private LLM
  • Agent-Based Travel PlannerContextual itinerary generation with agents.Private LLM
  • AI Evaluation ToolStructured, explainable assessment of AI output.Private LLM

Advisory

Work with me.

I'm open to collaborations in four areas.

Idea to working prototype

You bring the idea. I turn it into something people can click through and react to, so you can test it with users, buyers, or investors before you commit a team.

Agentic AI architecture

Multi-agent systems with context memory, graph-validated reasoning, and tool orchestration, designed so the outcome is transparent and controllable.

Transformation and platform strategy

Modernization roadmaps, incremental or transformation-led, that balance business outcomes, cost, and long-term flexibility. Modular by design, to avoid lock-in.

Executive advisory and due diligence

An independent view on high-stakes architecture decisions, technical due diligence, and product acceleration.

Talks and workshops. I also run sessions for leadership teams on AI adoption, from a keynote to a full-day working session.

Experience

Inside the big firms. Inside the startups.

Over 23 years on both sides. I've watched from close range how products get built, how AI gets infused into them, and how large transformation deals are won.

Global enterprises

HCL · Infosys · Mindtree · BNY Mellon · HTC Global Services

Scale, governance, and multi-million-dollar programs across technology services and banking.

Startups

Neeyamo · CoreStack

Small teams, where I saw product decisions made and shipped at close quarters.

How products get built

From first architecture to release, alongside product and engineering teams. I still build my own, so the view is current.

How AI gets infused

Taking AI from a pilot to a working part of the product: agents, retrieval, and document intelligence, on private and hosted models.

How large deals are won

Shaping the solution, the estimate, and the executive story for multi-million-dollar transformation pursuits. I wrote a book from that experience, The Solution Architect's Journey: Mastering RFP's through stories.

Now

Strategic solutioning for enterprise transformation · AVP at a top global consulting firm

I shape the solution for top-tier enterprises running multi-million-dollar transformation programs. What I bring is the ability to read a deal across its dimensions early: the business case, the architecture, the commercials, the delivery risk, and the people who have to say yes.

The work centres on finance operations (Accounts Payable, Procure-to-Pay, Record-to-Report), with agentic AI and SAP at the core.

Author

The Solution Architect's Journey: Mastering RFP's through stories

An ebook for solution architects on winning RFPs, told through stories.

AI work

  • Architected a multi-agent AI platform that combines LLM orchestration, knowledge graphs, and dynamic tool selection, used for enterprise configurators and AI copilots.
  • Designed two reasoning modes, guided and expert, on one adaptive state machine.
  • Designed agentic Accounts Payable solutions that pair deterministic workflows with LLM reasoning for exception handling and decision support.
  • Delivered internal AI platforms on private LLMs: code assistants, retrieval-augmented chatbots, and document intelligence pipelines.
  • Advised on multi-pass LLM extraction across 7 financial document types, backed by 3,100+ automated tests.
  • Set up an AI centre of excellence to tie innovation back to business outcomes.

Transformation

  • Led global, multi-million-dollar transformation programs.
  • Modernized Java, Spring Boot, and Azure estates, with 30–40% efficiency gains from modular architectures.
  • Ran an architecture centre of excellence and built a factory-based 6R migration framework.
  • Set up a FinOps centre of excellence for cost accountability across business units.
  • Enterprise & solution architecture
  • Agentic AI systems
  • LangGraph · LangChain
  • State-machine design
  • AP automation · P2P · R2R
  • SAP S/4HANA · BTP · VIM
  • Intelligent document processing
  • Cloud-native & microservices
  • Data platforms & observability
  • Presales & RFP solutioning
  • Architecture governance

Blog

Notes from the work.

Agentic AI in the enterprise, finance operations, and how large deals are won.

Build software around intentions, not applications

Nobody wants to use an application. They want something done. Agentic AI finally lets us design for that.

2 min read

Let the model propose. Let the state machine decide.

Agentic AI demos tend to break in the same places once they reach production. The fix is an old idea, deterministic control around the reasoning.

2 min read

Read the deal before you write the solution

A large RFP response is often decided before anyone draws the architecture. Five things to read early.

2 min read
All posts More on LinkedIn ↗ MyTechDives, my earlier blog ↗

Contact

Let's find the right starting point.

Pick the kind of session you have in mind. Every request comes straight to me, and I reply within one to two business days.

Connect on LinkedIn ↗

Based in Chennai, India. Away from work I'm usually planning the next road trip.