Applied AI and Data Engineer
I build machine-learning services, data platforms and decision products across Python, TypeScript and cloud infrastructure. Currently completing a Master of Data Science and Decisions at UNSW. Full working rights in Australia.
Featured system / Finlo
A three-service product combining scenario modelling, financial rules and controlled ML workflows.
Read the case study ↓Decision surface
Repayment decisions across currencies depend on income timing, exchange-rate movement, available cash and financial risk. These factors are normally viewed separately.
Finlo brings scenario modelling, repayment readiness, FX awareness, planning helpers and model-assisted recommendations into one controlled decision workflow.
Architecture evidence
Open a concise, public-safe explanation of each design boundary.
Engineering judgement
Each project is presented through the problem, architecture and operational constraint it addresses.
Infrastructure report
Marketing and content intelligence across application, data, NLP and infrastructure layers.
Terraform / ECR / CI-CD / cost review / service health
React and Vite drive the interface; Express, PostgreSQL and Drizzle hold the application layer; Redis-backed workers separate queued workloads.
Python NLP services support RAG and sentiment workflows. Terraform, ECS, RDS, ALB, ECR and CloudFront define the deployment boundary.
Investigative technology feature
A local-first system built around dry-run defaults, claim-to-evidence enforcement, privacy scanning, approval gates and an auditable path from preparation to action.
Urban reasoning / Mobility QA
Academic research engineering across benchmark construction, geographic and temporal reasoning, question design, evaluation and group research workflow. View repository ↗
Representative benchmark format“Which location follows the final observed stop, and what temporal evidence supports that answer?”
Illustrative public-safe question structure; not a disclosed benchmark item.
File 05 / Evaluation report
The reproducible evaluation covers deterministic extraction across diverse skill packs.
File 06 / Performance evidence
A 100,000-row enrichment flow completed across three repeatable runs with zero row loss and a 4.16-second median.
View evidence ↗Blockchain / application security / NLP / firewall policy
Hybrid consensus research balancing computational competition, contributed resources, governance and privacy in a distributed system.
A combined framework for ML-based XSS detection, HTTP-header assessment, SQL-injection protection and network-level controls.
Mixed-method analysis of early public discussion around large language models, combining sentiment labels, topic discovery and qualitative interpretation.
A literature review spanning policy anomalies, distributed firewall configuration, automation and formal approaches to security rules.
Titles and dates are held as archive facts, without auto-calculated duration labels.
Sydney
Building a cross-border repayment intelligence platform focused on explainability, decision support and model-backed scenario analysis.
Remote, Part-time
AI research, RAG systems, strategic risk intelligence and production-oriented development.
Sydney
Role matching, data onboarding, evaluation tooling and recommender-system workflows.
Sydney
Market intelligence, company research and decision-focused financial reporting.
Texas, Remote
RAG architectures, recommendation systems and enterprise LLM workflows.
Bengaluru
ML-driven detection of XSS and SQL-injection patterns with application-security analysis.
A production system is not properly designed when reliability depends entirely on spending more.
Useful automation can remain meaningfully constrained when consent and evidence are designed into the state machine.
A recommendation has limited value when the assumptions, trade-offs and uncertainty remain hidden from the person acting on it.
Technologies are indexed to work where they were applied.
Formal study indexed to the research and systems work recorded in this edition.
Record 01 / Postgraduate
Causal inference, game theory, econometrics, machine learning and distributed systems.
Record 02 / Undergraduate
Data structures, AI and machine learning, databases and computer networks. Best Project and Merit Award.
Open to serious engineering conversations
Based in Sydney with full working rights in Australia. Open to applied AI, machine-learning engineering, data engineering and technically demanding product roles.
dhaneshrameshofficial@gmail.com ↗Observed Income timing and the planned repayment overlap.
Change Test one assumption while holding the remaining inputs stable.
Trade-off Earlier repayment reduces exposure but narrows available cash.
Limit This is scenario support, not financial advice.
One variable can be changed at a time so the user can see which assumption produced a different result.
The simulator is a decision surface, not a prediction theatre.
Protected product routes use short-lived access state while refresh handling remains a separate controlled path.
The public abstraction records responsibility boundaries without disclosing secrets, internal identifiers or deployment configuration.
Deterministic calculation, explicit assumptions, stable explanation fallback.
Controlled model inference adds contextual comparison without bypassing product rules.
Keeping the paths separate makes capability, cost and failure behavior observable.
Training and evaluation concerns stay outside request-time execution.