Syncing Neural Context
Vansh Kamra
000 / 100 SYSTEM.ACTIVE
[ Portfolio · 2026 ] v2.0 // signal_active

Building
Scalable
Intelligence

· CS @ Scaler School of Technology

Vansh Kamra — building ML and LLM systems: QLoRA fine-tuning, LangGraph agents, and production MLOps pipelines from streaming ingestion to drift detection.

Explore Field
Backend · Distributed Systems · AI Infrastructure · ◆
Backend · Distributed Systems · AI Infrastructure · ◆
· 02 — Manifesto

Beyond
The
Code

I build systems that are observable, correct under load, and not painful to debug at 2am.

Microservices Built
15+
Fraud Scoring p95
<10ms
Ascent Hackathon
1st
· 03 — Records

Field Artifacts

C
Project_01
LLM Fine-Tuning / Security AI

CYBERTHREAT
INTEL

Phi-3 fine-tuned with QLoRA on NVD CVE data. MITRE ATT&CK mappings, YARA + Sigma rules. Live demo on Hugging Face.
F
Project_02
MLOps Pipeline

FRAUD
SHIELD

Kafka + FastAPI scoring at <10ms p95. Strict temporal split, ROC-AUC 0.907. Airflow retraining, drift detection.
I
Project_03
API Infrastructure

INFER
GATE

5-microservice API gateway. Redis rate limiting, distributed tracing, model routing.
G
Project_04
Applied AI / Evaluation

GRADE
SENSE

Grades answer scripts against a marking scheme and annotates the page region holding the evidence. Every score traceable to source.
A
Project_05
DevTools CLI

CODEBASE
ANALYZER

Extracted 31k+ functions, 182k call-graph edges, complexity scores from Django-scale Python codebases.
D
Project_06
RL / SQL Environment

DATAWAREHOUSE
OPS

Gym-compatible RL environment for training data-engineering agents. Three task tiers with shaped reward functions.
· 04 — Signal ·

Sync the
Signal.

Direct
vanshkamra27@gmail.com
Connect
LinkedIn
Explore
GitHub