$ Faculty of Technology · University of Delhi

Ashish Pal.

full-stack · ai/ml · systems

Full-stack engineer & AI builder — shipping systems
from legal analysis vectors to real-time WebSockets frameworks.

view_work() → get_in_touch()
scroll
Ashish Pal
AP
about_me

Who I Am

I'm Ashish Pal, a 3rd year Computer Science student at the Faculty of Technology, University of Delhi. I build highly decoupled full-stack products, specialized AI infrastructure pipelines, and perform algorithmic systems research.

My development experience ranges from engineering a Top 10 AI solution out of 270,000+ teams globally at the Google Cloud Gen AI Exchange Hackathon, to co-authoring comprehensive diagnostic neural network pipelines. I filter heavily for system ownership and real-world production optimization.

Currently focusing on low-latency microservices, context grounding (RAG), and transactional security architectures. Read my research manuscript →

Top 10
google_cloud_finalist
10
shipped_systems
DU
university
selected_work

Projects

01 / 10

LegalSummary AI

AI-powered multi-clause legal document contract parsing system cutting manual analysis duration by 80%. Integrates explicit context containerization hosted across GCP environments, combining modular file splitting with Redis caching abstractions.

next.js 14 typescript redis docker gcp cloud run firebase
↗
02 / 10

MeetWise

Real-time interview coordination application delivering collaborative sandboxed runtime programming instances sync'd via Convex WebSockets under sub-50ms thresholds. Integrates Native Stream video configurations, Monaco state binding, and automated recording hooks.

next.js convex websockets clerk auth stream video sdk monaco engine
↗
03 / 10

CredexAudit

Client-side cost evaluation parser auditing computational tool overhead configurations under two minutes. Utilizes localized storage vectors with tokenization summaries triggered at 2200 tokens/s via Cerebras Meta Llama instances and strict deterministic safe templates.

next.js 16 (app router) typescript strict tailwind v4 supabase postgres cerebras api vitest
↗
04 / 10

Research Lens

Asynchronous corporate literature graph engine mining knowledge omissions from text corpora. Implements heavy mathematical multi-dimensional parsing arrays mapped via FastAPI routines using spaCy, UMAP, and HDBSCAN token mapping clusters.

python backend fastapi spacy nlp umap clustering hdbscan knowledge graphs
↗
05 / 10

GolfStake

Asynchronous multi-role subscription orchestration dashboard supporting separate cryptographic access spaces via PostgreSQL Row-Level Security parameters. Connects production runtime validation via transactional Stripe checkout webhooks.

next.js core node.js postgresql strict supabase rls stripe webhooks
↗
06 / 10

PebloNotes

Decoupled collaborative notes application and workspace organization utility tracking real-time layout structures. Anchors secure data mutations across an independent FastAPI execution block connected to structured schema relational tables.

react.js typescript fastapi server python postgresql docker Compose
↗
07 / 10

Insight Synthesizer

Automated multimedia pipeline generating downloadable tokenized narration tracks from unformatted corporate text files and multi-page PDFs. Tracks data breakdown components via PyPDF text filtering rules and NLTK text sentence division.

python core nltk hugging face transformers elevenlabs tts node.js
↗
08 / 10

Vedaz-Booking

Full-stack dynamic calendar sync portal managing complex specialist advisory assignments and patient listing flows. Runs state distribution routines using explicit web infrastructure routes paired with non-relational storage collections.

react frontend node.js runtime express.js mongodb clusters rest webhooks
↗
09 / 10

KoinX-Backend

High-throughput accounting verification microservice parsing unstructured computational records and large ledger data documents. Maps array reconciliation matrices across heavy file streams to match asset calculations without latency bottlenecks.

node.js microservices express.js mongodb csv tracking streams data pipelines
↗
10 / 10

Pneumonia Detection CNN

Co-authored diagnostic academic computer vision paper assessing five distinct convolutional architectures over 26,684 image instances. Implements explicit Grad-CAM visual layers tracking network feature activation boundaries.

python research pytorch core grad-cam visual deep learning cnn layout
↗
expertise

Skills

languages
javascript (es6+)
typescript
python
c / c++
frontend & fullstack
next.js / react
tailwind css
node.js / express
convex / redis
mongodb / postgresql
ai / ml & infra
pytorch
hugging face
llm apis
docker / gcp
git / github
contact

Let's Build
Something.

Open to collaborations, internships, and interesting conversations. My inbox is always open.

ashishpal2804@gmail.com
github linkedin leetcode