Skip to content
View Swakie35's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report Swakie35

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Swakie35/README.md

Hey, I'm Swaraj 👋

Applied AI & Backend Engineer · Python / Java / SQL · MSc Computer Science, University of Liverpool

I'm a software engineer who enjoys the work between a promising prototype and a system you can actually reason about: clean APIs, dependable data pipelines, measurable ML results, useful observability and sensible failure handling.

My background spans software, data and applied AI work in India and the UK, including Python/SQL workflows, identity and access management, retrieval-based systems and operational analytics. I completed my MSc in Computer Science at the University of Liverpool in 2026. Lately, I've been building more deeply around agent tooling, backend reliability and evaluation-driven ML.

What I work with

Area Technologies and interests
Backend & systems Python, Java, FastAPI, REST, SQL, PostgreSQL, Redis, asynchronous processing
AI & ML scikit-learn, gradient boosting, forecasting, NLP, retrieval, agent workflows, evaluation
Infrastructure AWS, Docker, GitHub Actions, CI/CD, logging and monitoring
Data Pandas, NumPy, SQL analytics, data modelling, pipeline design

Selected engineering work

I use these repositories to explore specific engineering trade-offs. They are self-contained reference implementations with synthetic or demonstration data, not claims of production deployment.

Repository Engineering focus
Switchboard Bounded read-only tool execution, evidence traces and optional LLM function calling
ClauseLedger Contract evidence retrieval with citations, a FastAPI interface and audit records
Counterflow Leakage-aware demand forecasting with time-ordered evaluation and baseline comparisons
QueuePilot Durable background jobs, retries, leases, idempotency and dead-letter handling
Keyline Java HTTP service for order lifecycles, optimistic locking and idempotent creation
QueryAtlas Executable SQL warehouse, incremental loading and analytical reporting

Also exploring: Driftwatch (ML drift monitoring) · Tracecraft (concurrency and telemetry) · TriageRank (synthetic research-only risk modelling)

Earlier work

My Path-Visualizer, Stock-Forecasting and other repositories are part of the same learning and engineering journey. I keep the earlier work visible because it's useful to see how approaches change over time.

What I care about

I like code that explains its assumptions. I'd rather show the baseline, the failure case and the test than just a polished demo. Right now I'm particularly interested in retrieval evaluation, safe agent/tool interfaces, data-intensive backend services and practical MLOps.

Outside engineering, cricket has been a big part of my life; I've captained at state level in India.

Connect: LinkedIn · GitHub repositories

Pinned Loading

  1. clause-ledger clause-ledger Public

    Citation-first contract evidence retrieval using FastAPI, SQLite and inspectable lexical search.

    Python

  2. counterflow counterflow Public

    Demand forecasting with leakage-aware feature engineering and walk-forward model evaluation.

    Python

  3. keyline keyline Public

    Java 21 HTTP order lifecycle service with optimistic locking and idempotent creation.

    Java

  4. Path-Visualizer Path-Visualizer Public

    Python pathfinding visualisation exploring graph-search algorithms, including A*.

    Python

  5. queryatlas queryatlas Public

    Executable SQLite analytics warehouse with star schema, incremental upserts and window-function reports.

    Python

  6. switchboard switchboard Public

    Step-bounded read-only agent runtime with validated tools, evidence traces, FastAPI and optional LLM function calling.

    Python