Focus mode
28 icons
All stack, no filter.
Pick a category to reshuffle the globe and surface the parts of the stack that matter most for that lane.
Syed Anees Ashraf.
I
B.Tech CSE @ Amrita (2027) building multi-agent AI systems, real-time collaboration engines, and scalable web architectures.
10+
Projects
15+
Technologies
1st
AI-Verse Hackathon
01
who am i
Crafting
meaningful
experiences.
A Computer Science student passionate about building intuitive and performant applications. I enjoy creating products that balance clean engineering with thoughtful AI/ML integration.
When not coding, I'm exploring emerging technologies, experimenting with design systems, or diving deep into agentic AI patterns and interactions.
Currently seeking my first developer role — an opportunity to learn fast, contribute meaningfully, and grow as an engineer.
Download Resume
Open to work
anees@portfolio:~/achievements
LIVE
# Amrita Vishwa Vidyapeetham · B.Tech CSE · 2027
$ run achievements --filter all
WINNER
AI-Verse Hackathon
200+ teams
FINALIST
OLabs · Gov. of India
Top 8 National
LEAD
Amrita IETE-SF CP Team
Technical Lead
Education
Amrita Vishwa Vidyapeetham
B.Tech CSEExpected 2027Coimbatore
02 — what i know
Skills & Technologies
Hover an icon
Python
A living map of the stack behind the projects in this portfolio.
01
Languages
Core languages behind every pipeline, API, and interface I ship.
02
AI / ML & Agents
Model training, retrieval, vision, and multi-agent orchestration.
03
Backend & Systems
APIs, data, cloud infra, and the dev tooling that ships it all.
04
Frontend & Mobile
Interfaces and cross-platform apps, from markup to component systems.
03 — what i've built
Selected Work
RAG Pipeline · 6 Agents
Query Decompose
→
Retrieve
→
Rerank ⚡
91+ VCs matched
87% top-3 acc.
FindIn'.ai
🏆 1st Place · AI-Verse Hackathon (200+ teams)Architected a 6-agent RAG pipeline — query decomposition → retrieval → reranking → summarization → VC matching → translation — over 3,292+ documents. Investor-matching engine ranked 91+ VCs via Cohere embeddings, achieving 87% top-3 accuracy on a held-out eval set.
6Agent pipeline
87%Top-3 accuracy
<2sE2E latency
CoherePineconeGroqFastAPIReact
REQ-042
Authentication
+ OAuth2 SSO support added
~ Session timeout: 30m → 15m
− Basic auth removed
2 editors live
🔒 Optimistic lock
SpecSync
MERN requirements management platform with RBAC (Admin / Editor / Viewer), Git-style versioned diffs, and WebSocket real-time collaboration. REST API uses optimistic locking to handle concurrent writes with signed-URL file uploads.
RBAC3 permission tiers
WSReal-time collab
GitVersioned diffs
MongoDBExpressReactNode.jsJWT
Training · Epoch 47/100
BLEU-40.31
ROUGE-L0.43
KL β0.82
KL annealing
T-CVAE + Mask R-CNN regions
Backstory Gen · T-CVAE
Designed a Transformer-based Conditional VAE conditioned on Mask R-CNN region features to generate pre-event narratives. Applied KL annealing to prevent posterior collapse — BLEU-4: 0.31 / ROUGE-L: 0.43 on the VisualStoryTelling benchmark.
0.31BLEU-4
0.43ROUGE-L
KLAnnealing applied
PyTorchMask R-CNNHuggingFace
Aisle 3 · ±0.8m accuracy
"blue running shoes"
Nike Air Zoom · 94%
Adidas Ultraboost · 88%
Collab filter
precision@5: 74%
Smart Shopping Assistant
BLE trilateration + Kalman-filtered indoor positioning achieving sub-1m accuracy. Integrated CLIP-based visual product search and a collaborative filtering recommender (precision@5: 74%) with real-time Firebase sync.
<1mIndoor accuracy
74%precision@5
CLIPVisual search
React NativeFirebaseBLECLIP
04 — let's connect
Get In Touch
I'm actively looking for my first developer role! Whether you have an opportunity, a project idea, or just want to say hi — my inbox is always open.
syedaneesashraf246@gmail.com