Kamal Mohan — AI Native Software Engineer & Full Stack Tech Lead

Profile

I've been building enterprise software since 2012, mostly in C# and .NET. For the last few years most of my work has been putting LLMs inside systems people already use every day, not building demos next to them.

At Bromcom I led the team that added natural-language reporting and an MCP-based assistant to a school MIS used in the UK. Before that I built a payment gateway at Pine Labs, moved a German healthcare retailer off a monolith, and spent time on-site in China shipping a manufacturing system.

Experience

Bromcom United Kingdom Remote

Dec 2022 – Aug 2026

Tech Lead

Natural-language report generator

Added a GPT-4o-mini assistant to the legacy C#/ASPX report builder. Staff describe the report they want in plain English instead of clicking through the field tree.
The prompt carries the MIS schema, so the model only picks fields and filters that exist. A typical request comes back in about 15 seconds.

.NET 8 ASP.NET Web API Dapper SQL Server GPT-4o-mini

MIS assistant on Model Context Protocol

Designed the assistant around MCP. The model calls typed MIS tools, gets live data back, and the answer streams to the browser over SSE.
Built the AskToMcpChatBot API and wired in a Python code interpreter. The model writes and runs its own analysis code, checks the result, and can hand back an Excel file.
Led the frontend, backend and DevOps people on the project, and owned the architecture and code reviews.

.NET 8 MCP GPT-4o Python SSE streaming

Pine Labs India

Jan 2022 – Nov 2022

Tech Lead

Plural payment gateway

Built a gateway that sits in front of several payment aggregators and routes each transaction. Also built Plural Console, the merchant dashboard for analytics and routing.
Implemented 3DS 2.0 authentication and tokenised card payments with OTP authorisation.
Moved the service from a single monolith to a multi-node setup so one node going down no longer took payments with it.

REST APIs Microservices Multi-node 3DS 2.0 Card tokenisation OTP

Fintech RAG assistant

Built a retrieval-augmented chatbot for financial advisory questions, covering ingestion, retrieval and answer generation.

FastAPI Pinecone LangChain Llama 3.1 70B MLflow W&B

ACL Digital client in Germany

Apr 2019 – Dec 2021

Tech Lead · Paul Hartmann hospital retail platform

Led the move from a monolith to service-based microservices on ASP.NET Core and Cosmos DB.
Rewrote a slow database job that took about 8 hours. It now finishes in around 2 minutes.
Worked with the client's enterprise architects in Germany on the design for EU deployment.

ASP.NET Core Cosmos DB Microservices SOA

Athena Technology Solutions clients in USA China

Mar 2016 – Jan 2019

IT Manager · Miasolé manufacturing execution system

Moved legacy ASP WebForms apps to MVC and shipped multilingual builds for plants in the US and China.
Led on-site delivery at Hanergy in China, and built performance-test tooling to prove the system would hold up at scale.

ASP.NET MVC ASP WebForms Multilingual (i18n) Performance testing

Tata Consultancy Services client in USA

Jul 2015 – Feb 2016

Senior Software Engineer · Citi ATM platform

Owned risk identification and mitigation for an ATM platform running in several regions.
Delivered multilingual modules rolled out across Citi's US ATM network.

Multilingual (i18n) Risk management

Earlier roles

2012 – 2015

Athena Technology Solutions, Software Engineer (2014–15). Hadoop work with Hive, Pig, MapReduce and Impala, mostly large batch processing and cluster tuning.
Avolot Technologies, Software Engineer (2012–14). User and content management modules, from functional design through deployment.

Hadoop Hive Pig MapReduce Impala

Portfolio

Four systems I designed and built. Each diagram shows how the parts talk to each other in production. The screens are rebuilt with sample data, because the real ones show client and student records.

MIS assistant on Model Context Protocol

Bromcom, UK · remote · 2023 – 2026

School staff ask questions about their own data in plain English. The model decides which MIS tools to call, the MCP server runs them against live data, and a Python sandbox does the analysis when a question needs more than a lookup. The answer streams back as it's written.

.NET 8 MCP GPT-4o SSE Python SQL Server

Architecture numbers follow one request

Output screen sample data

MIS · Ask a question

Which Year 9 forms have the most students under 90% attendance this term?

✓ get_attendance(year: 9, term: "Autumn")212 rows

✓ run_python(group by form, filter < 90%)0.8 s

14 students in Year 9 are under 90%. 9F has the most, with 5.

Form	Under 90%	Lowest
9F	5	81.2%
9K	4	84.0%
9R	3	86.5%
9T	2	88.1%

⤓ year9_attendance_autumn.xlsx

Want the list of names for 9F?

Natural-language report generator

Bromcom, UK · remote · 2023 – 2026

A plain-English front end for a reporting module that has run in production for years. The prompt builder sends the schema with the request, the model returns a report spec, and the API checks it before anything touches the database. The old report viewer shows the result, so nobody had to learn a new screen.

.NET 8 Web API Dapper ASPX GPT-4o-mini SQL Server

Architecture request loop

Output screen sample data

Reports · Describe your report

Pupils with more than 3 behaviour incidents this half term, with tutor group and SEN status

Build

Read as: NameTutor groupSEN statusIncident count filter incidents > 3 this half term

Pupil	Tutor group	SEN	Incidents
A. Patel	10B	K	7
J. Morris	9F	None	5
L. Okafor	11C	E	5
S. Nowak	8A	None	4

4 rows · generated in 14 s · Save as template · Export CSV

Fintech RAG assistant

Pine Labs, India · 2022

Answers advisory questions from internal financial documents. Ingestion runs offline: documents are chunked, embedded and stored in Pinecone. At question time LangChain pulls the closest chunks, Llama 3.1 70B writes the answer, and MLflow and Weights & Biases keep track of which prompt and model version produced it.

FastAPI LangChain Llama 3.1 70B Pinecone MLflow W&B

Architecture ingestion + query path

Output screen sample data

Advisory assistant

Can a merchant on the Growth plan get settlement faster than T+1?

Yes. Growth plan merchants can request same-day settlement once they have 90 days of transaction history. A fee applies on each settled amount, and the request has to be approved by the relationship manager.

settlement_policy.pdf · §4.2 growth_plan_terms.pdf · p.3

What's the fee?

0.15% of each same-day settled amount, billed monthly.

Plural payment gateway

Pine Labs, India · 2022

One API for merchants, several aggregators behind it. Every payment goes through 3DS 2.0, cards can be tokenised for repeat use, and the routing engine picks the aggregator per transaction. Merchants see all of it in Plural Console.

Microservices REST APIs Angular Multi-node Txn store

Architecture payment path

Output screen sample data

Plural Console · Transactions

Today's payments

AllFailedRefunds

Txn	Amount ₹	Method	Routed to	Status
PL-48213	2,499	Card · 3DS	Agg. A	Captured
PL-48212	860	UPI	Agg. B	Captured
PL-48211	12,000	Saved card	Agg. A	OTP sent
PL-48210	1,150	Netbanking	Agg. C	Declined
PL-48209	399	UPI	Agg. B	Captured

Export CSV · Routing rules · Settlement report

Where I've delivered

Profile

Experience

Portfolio

MIS assistant on Model Context Protocol

Natural-language report generator

Fintech RAG assistant

Plural payment gateway