Noida, India
Abhinav Dwivedi
Software engineer. Data platforms at HCLTech. Security systems and product software in the open.
By day I migrate Matillion and Snowflake pipelines. Otherwise I build host DNS security and post-quantum tooling.
Noida, India
Software engineer. Data platforms at HCLTech. Security systems and product software in the open.
By day I migrate Matillion and Snowflake pipelines. Otherwise I build host DNS security and post-quantum tooling.
Each frame is a system. Click it for the longer plate: the problem, the build, and a drawing of how it moves.
Phosphor intercept
An on-host DNS filter. Known-bad names are blocked before a connection opens.
The query is caught on the machine, across UDP, TCP, and DoH.
A local blocklist, then entropy, then the wording of the name. Nothing leaves to make that call.
Block, ask, or allow before a socket opens. The core is Go. The Mac client is Swift.
A program resolves a name, then opens a socket. If the decision happens after that, the connection already exists. The useful place to stop a known-bad domain is on the host, before the lookup leaves.
Lattice beside the old lock
Proxy and certificate tooling for connections that need a post-quantum path.
Find RSA and elliptic curves still in use, so the migration has a list instead of a slogan.
A proxy speaks the older handshake on one side and a hybrid post-quantum handshake on the other.
A certificate authority adds ML-DSA beside the classical signature. Services are Go. The trust center is Next.js.
Recorded traffic that uses RSA or elliptic curves can be read later, once a capable quantum computer exists. Replacing every client at once is not a real plan. The edge can move while older clients still speak.
A warm room, unfinished
A cross-platform social app for considered introductions. Still being built.
Android, iOS, and web, meant to feel like the same product.
A Go service behind the clients. Matching and conversation are still being shaped.
Codex writes test scripts. Gemini models run those tests. A coach that calls a model for members is not what the app does today.
Most social products optimize for volume. Kindred is aimed at fewer, deliberate connections, with the member still in control of the pace.
A clinical ledger
A Spring Boot service that checks a claim before it is treated as payable.
A claim arrives as JSON over HTTPS into a controller that validates the shape first.
Pluggable checks for duplicate submissions, fee discrepancies, and member eligibility.
Findings land in PostgreSQL through parameterized queries. Java 17, Spring Boot, JUnit.
In a US payer system, a claim that skips these checks is how duplicate billing and bad codes get paid. The engine is the gate, not the clinic.
A current, not a batch
Change data from the source, through a stream, into tested models.
Debezium reads the database log so a row change becomes an event.
Redpanda holds the stream. The warehouse is not polling the source.
dbt builds the models and tests. OpenLineage keeps the trail from source to table.
The point is a path you can inspect: what changed, where it went, and which model consumed it.
A notebook that runs
Daily Python practice in the browser, with the runtime kept off the main thread.
Monaco is the editor. A pool of Python basics exercises sits behind it.
Pyodide loads in a web worker. Syntax errors, assertion failures, stdout, and a five-second timeout come back to the page.
Tutor chat tries Gemini first and falls back to Groq. The key stays in a Cloudflare Worker, not in the page.
Daily drill progress lives in localStorage. This one is a local project, not a public repository yet.
A yellow cap
A macOS menu-bar app and a launch daemon that hold the battery at a chosen limit.
A slider sets the ceiling, for example 80 percent, and the daemon applies it.
The machine can drain toward that limit even while the adapter is plugged in.
An assertion keeps the Mac awake for that discharge in clamshell, without an external display.
The SwiftUI app talks to the daemon over a local channel. The daemon is the part that touches power. This is a local project.
Paper, then light
A WebGL promo and a generated hardcover for Between Us: Dad’s Money Talks, by Neeraj Bharti.
A Python script builds a glTF book: cover, spine, page block, and textures a 3D tool can open.
The browser scene lights that book and plays a cinematic sequence.
A product demo for a book, not a storefront. Local project.
The asset is small enough to drop into Blender, Unity, or the web. The page is the showing of it.
An archive with a draft
An audit agent for the HackerRank Orchestrate challenge. A deletion marked complete can still leave a person behind.
First decide whether deletion was actually requested. Otherwise the status is not applicable.
Databases, logs, attachments, exports, search indexes, archives, backups, and restores that bring a row back.
Ordered remediation and verification steps, not a single boolean.
The solution reads the challenge dataset and writes a decision per case. It is a challenge entry, not a product repository.
HCLTech · Noida, India
Developed and tested the migration of 100+ pipelines from IBM DataStage to Matillion and Snowflake, in a team of six. Automated source-to-target checks on 50+ tables caught defects before release.
I administer the Matillion and DataStage platforms: debugging sessions, developer errors, and Grafana. With the DevOps team I use Azure DevOps pipelines and automated Dev, QA, and Prod releases so those handoffs are not manual.
Python scripts group recurring failures in Unix job logs. Troubleshooting a repeat failure went from about 3 hours to 30 minutes.
HCLTech · Noida, India
Trained in QA and Java test automation: writing automated test cases, running them, and reporting defects.
Samsung R&D Institute Bengaluru · Remote
Built Python services in Django and Flask, and machine learning models on IoT sensor data, at 99% accuracy. Samsung PRISM awarded a Certificate of Excellence and an internship certificate.
A. Dwivedi, G. K. Saini, U. I. Musa. 2023 2nd International Conference for Innovation in Technology (INOCON).
A. Dwivedi, K. Sood, N. Aggarwal, H. H. J. Sapingi.
maintainer I maintain the open-source repositories for Terminator Sec and QuantumShield.
Python (FastAPI, Flask, Django), Java, Go, SQL, Bash.
Azure DevOps CI/CD, Git, Linux/Unix, test automation, debugging, DevSecOps training.
Matillion, IBM DataStage, Snowflake, Grafana, ETL pipelines, microservices, REST APIs.
LangChain, MLflow, SageMaker, Databricks, Vertex AI, Azure ML, ArgoCD.
Chandigarh University, Mohali, India. 2021–2025.
Coursework at HCLTech.
Agent Development and Deployment.
I read these. Email also works if a form is the wrong shape: abhinavd372@gmail.com.