Portfolio / 2026 Bengaluru, India
00 / Positioning

Vidwan Gowda H M — Product-minded Software Engineer, Bengaluru Deciding what to build and building it are the same job.

I'm a software engineer at Oracle working on telecom order orchestration. Last year I found a problem nobody had filed a ticket for — engineers burning two hours per incident tracing failed orders by hand. I ran the interviews, wrote the spec, cut the scope, built the system, and deployed it to all fifteen engineers in the target QA function. Triage now takes twenty-five minutes.

The product half

Discovery interviews, PRDs and acceptance criteria, ruthless prioritisation, phased rollout, adoption I actually had to earn. One product, owned end to end, in active internal use across the target QA function.

The engineering half

Python is where I'm fastest; Java, XQuery and SQL are what the job runs on, across six integrated BSS/OSS systems. In a domain where a mishandled retry means a customer's service never activates, I design for the failure case first.

80%cut in fault triage time
15hengineer-hours returned weekly
15/15adoption across target QA function
200+students on a govt-funded platform
6downstream systems integrated
Open to APM / Product Manager  ·  SDE / Backend Engineer  ·  Bengaluru
01 / Now

Software Engineer, Oracle

Aug 2025 — Present

Communications: Order & Service Management

I deliver Oracle OSM solutions for telecom fulfilment — orchestration, provisioning, activation and task flows across five workflow domains for enterprise carrier customers. I write the orchestration in Java, XQuery and SQL, integrate six downstream BSS/OSS systems (UIM, ASAP, OSB, ERP, AIA, BRM), and harden the retry and failure-recovery paths under production volumes.

The other half of the job is translation: turning business requirements into specifications, configurable workflows and test scenarios, and owning SIT/UAT strategy and root-cause playbooks across five stakeholder functions — QA, product, integration, upstream and downstream engineering.

JavaXQuerySQL RESTJMSOracle OSM SIT / UAT ownership

Previously: Project Intern at Oracle (Jan — Jul 2025) — configured and validated OSM orchestration across two enterprise telecom domains, and authored the SIT/UAT scenarios, defect replays and edge cases that went with them.

02 / Case study
0-to-1 internal product · sole owner

OSM_Vassist

Nobody asked for this. I found the problem in user interviews, quantified what it was costing, and then did every part of the job — research, PRD, prioritisation, architecture, implementation, rollout, adoption tracking.

Before

~2 hours per incident, manually walking COM, SOM and TOM layers to find which of six downstream systems had thrown.

After

25 minutes, with each supported failure class classified and routed to the team that owns it — automatically, with evidence attached.

  1. 01 — Discovery

    I interviewed QA and support engineers about their week, not about features. The same complaint surfaced repeatedly and had never been filed: tracing a failed order was manual archaeology across three order layers. I timed it, multiplied it by incident volume, and put a number on the recurring cost — which is what made the case for building anything at all.

  2. 02 — Spec

    I wrote the PRD: acceptance criteria, edge cases, and the operational decision rules that determine who owns a given failure. Seven capabilities, scoped and sequenced — order lookup, fault-detail extraction, failure triage, repush eligibility, abort handling, skip/bypass provisioning, process-history reporting.

  3. 03 — Prioritisation

    I ranked capabilities by triage frequency against time saved per incident, and cut the ones that wouldn't survive contact with a live support workflow. Several features I liked didn't make it, because they solved rare cases at the cost of the common path.

  4. 04 — Build

    Python. A capability router at the front, a rules engine that classifies each failure and maps it to the owning team, and structured JSON/CSV evidence generation with debug artefacts and verdict reports — so the output of a triage run is something you can attach to an escalation rather than retype.

    PythonCapability router Rules engineJSON / CSV evidence Oracle OSM APIs
  5. 05 — Rollout

    Shipped as a versioned internal release with onboarding documentation and a phased rollout rather than an announcement. It reached 100% adoption across the target QA function — all fifteen test engineers spanning SIT and UAT — and reduced escalation churn, because faults stopped being passed between teams before landing on the right one.

2h → 25mmean triage time
~15 hrsengineer-hours returned per week
7capabilities shipped, sequenced
100%adoption, 15 of 15 engineers
What I'd fix

I tracked adoption by asking people. That worked at fifteen users and would have failed at fifty. The first thing I'd add is usage telemetry — which capabilities actually get run, which failure classes recur, where the rules engine returns a verdict nobody trusts. I shipped the product before I shipped the feedback loop.

03 / Selected work

Three projects, each with a decision behind it

Every one of these had a constraint that determined the whole design. The left column is the call I made; the right is what I built to honour it.

Quiz Wizards

Govt. of Karnataka funded

A camera-based interactive quiz platform, shipped to 200+ students across two government schools inside a fixed grant window.

The decision

Rural classrooms have no reliable internet. Rather than treat that as a degraded case, I made offline the default and designed a pen-drive content-sync journey so teachers could refresh question banks with zero connectivity. That single constraint drove every other choice.

The build

Automated setup, launch, backup and data operations so non-technical teachers could run it unaided. No on-site support, which is what made the rollout repeatable from one school to the next instead of a one-off deployment.

Source → Python · OpenCV · Offline-first

Smart Route

Live

Truck route planning and cost estimation that lowered modelled fuel spend, emissions and operating cost by 7–10% per route.

The decision

Shortest route isn't cheapest route. Fleet operators care about landed cost per trip, so I made the output a cost figure rather than a distance — tolls and fuel estimated by vehicle class, not averaged.

The build

Bing Maps API for route visualisation and ETA prediction, with a toll and fuel model layered on top. Cost estimates landed at 80% accuracy against actuals.

Live app → Python · Streamlit · Bing Maps API

Finsight AI

Live

Real-time market intelligence over 1,000+ live equity and cryptocurrency data points.

The decision

A chart tells you what happened; it doesn't tell you what to make of it. I used the model for interpretation on top of real market data — never for the numbers themselves, which stay sourced and verifiable.

The build

Python and Streamlit, pulling Yahoo Finance market data and passing structured context to Google Generative AI for tailored per-company insight.

Live app → Python · Streamlit · Gemini · yfinance
04 / Archive

Smaller builds, kept for the range they show rather than the depth.

MusifyFull-stack music app — auth, search, trending HR Analytics DashboardPower BI — performance, diversity, turnover General Elections 2019Power BI — constituency-level election metrics CiniMagicRecommendation engine — Python, pandas Bank Smart ContractSolidity, Ethereum Blind Assist 2.0Vision-language model for navigation assistance Fertilizer PredictionRandom Forest, KMeans World Metro AnalysisPython, Plotly, Matplotlib Job Market ScraperPython, Selenium
05 / Capabilities

What I've used, and where I used it

No percentages. Each group carries the thing I actually shipped with it, so you can discount the rest accordingly.

Product
PRDs & specificationsAcceptance criteria Product discoveryBacklog prioritisation (RICE, MoSCoW) User stories & story mappingRelease planning Stakeholder managementJira · Confluence · Figma

Used on: OSM_Vassist end to end, and the requirement-to-specification work in my day job at Oracle.

Research & metrics
User interviewsJourney mapping Jobs-to-be-DoneSuccess metrics & KPIs Funnel analysisCompetitive analysis Power BI

Used on: the discovery interviews behind OSM_Vassist, and two published Power BI dashboards.

Languages & backend
PythonJavaSQL XQueryJavaScript / TypeScriptC++ REST API designFastAPI · Flask · Node.js JMS message queuesSystem design DSAUnit testing (JUnit, PyTest) OAuth / JWT

Used on: production orchestration across six integrated systems at Oracle; OSM_Vassist in Python.

Data, ML & applied AI
PostgreSQL · MySQL · MongoDBpandas · NumPy · scikit-learn LLM integration (Gemini, OpenAI, Claude)Prompt engineering Structured output (JSON mode)RAG & semantic search LangChainOpenCV · VLMs

Used on: Finsight AI, CiniMagic, Quiz Wizards' vision layer, Fertilizer Prediction. Honest caveat — my LLM work is application-level integration, not model training or production eval harnesses.

Cloud, DevOps & process
DockerKubernetesCI/CD · Jenkins AWSGit · code reviewLinux Agile · Scrum · Kanban
06 / Code

Practice, in the open

I practise algorithms regularly — mostly in Python and Java — and everything I build stays public.

31Public GitHub repositories

My current solved counts are on the LeetCode profile below.

LeetCode → GitHub →
07 / Education

B.E., Information Science & Engineering

2021 — 2025

Ramaiah Institute of Technology, Bengaluru  ·  CGPA 9.03 / 10.00

Certifications
AI Product Manager
IBM · Professional Certificate
Business Analysis Fundamentals
Microsoft
Oracle APEX: Foundations
Oracle
08 / Résumé

The short version

Vidwan Gowda H M
Updated July 2026 · 2 pages · PDF
  • NowSoftware Engineer, Oracle — Communications: Order & Service Management
  • OwnedOSM_Vassist end to end — discovery through rollout. Triage 2h → 25m, 15/15 adoption.
  • Builds inJava, Python, SQL, XQuery — across six integrated BSS/OSS systems
  • StudiedB.E. Information Science, Ramaiah Institute of Technology — CGPA 9.03
Download PDF ↓ Open in new tab →
09 / Contact

Tell me what your team is stuck on.

Product roles or backend engineering, in Bengaluru. Thirty minutes is usually enough to work out whether it's a mutual fit.

Book 30 minutes → Email me
Email
connect.vidwangowda@gmail.com
Phone
+91 79750 45560
LinkedIn
in/gowdavidwan2003
GitHub
gowdavidwan2003