Asmitha M

AI Engineer · Data Engineer · Full Stack Developer

Building production-grade intelligent systems — agentic AI, big data pipelines, edge inference, and reinforcement learning. 7 deployed systems across 8+ national and international hackathons.

Systems Built
7+
Hackathons
8+
CGPA
9.3
Records Processed
200K+

About Me

I am an AI Engineer focused on building production-grade intelligent systems. Bridging the gap between pure research and practical application, I design architectures that don't just train well, but deploy seamlessly. My work spans agentic workflows, deep reinforcement learning, and data analytics, backed by full-stack engineering expertise.

I believe true AI engineering happens beyond the model weights. It requires rigorous data engineering, robust backend infrastructure, and scalable system design. I don't just train models — I architect end-to-end autonomous systems that solve complex, real-world problems.

Projects

ROOT/RESEARCH/PROJECTS [7]
BIG_DATA: SUBINTEL

Subintel

Subscription Revenue Intelligence Pipeline

PySpark 3.5 Kafka PostgreSQL 15

Big data ETL/ELT pipeline ingesting 200K+ billing events across 500 SaaS tenants with MRR/ARR/churn analytics.

  • Year/month-partitioned Parquet storage with predicate pushdown at <50ms query latency.
  • Implemented multi-tenant row-level security and schema enforcement with dead-letter sink validation.
PySpark Kafka PostgreSQL Docker Compose CI/CD
LAKEHOUSE: SAAS_METRICS

SaaS Metrics Lakehouse

Event-Driven Medallion Analytics Platform

Bronze→Silver→Gold Delta Lake

Bronze→Silver→Gold medallion data lake design mirroring enterprise-grade sustainable technology data lake architecture.

  • ETL/ELT orchestration via Airflow DAGs (retries, SLA alerting) to star-schema PostgreSQL warehouse.
  • Idempotent incremental loads for live BI/reporting dashboards with centralized schema catalogue.
Apache Airflow Delta Lake Databricks Power BI PostgreSQL
INGESTION: SENTINEL_PRO

LLM Sentinel Pro

High-Throughput Ingestion Pipeline

5,000 rec/12s HIPAA Auditing

High-throughput secure ingestion pipeline optimized from nested vectorization bottlenecks to 417 rec/s via pre-caching.

  • Achieved 100% schema recall and 94.2% precision under HIPAA-compliant compliance-oriented audit schema.
  • Automated CI/CD suite with 9 unit tests passing in <1.5s.
Python FastAPI PostgreSQL Pydantic Docker
AETHERFLOW AI

SupportFloww

Confidence-Gated Support Intelligence

DistilBERT MC Dropout

Uncertainty-aware AI routing engine that detects ambiguous B2B SaaS tickets before misrouting.

  • 3-tier decision gate (Route/Clarify/Escalate) using Monte Carlo Dropout entropy.
  • Reduced unnecessary escalations by 71% and SLA breaches by 39%.
FastAPI Transformers XGBoost Python
INCIDENTMIND

IncidentMind

Autonomous Production Incident Resolution

Meta Finalist RL Research

Autonomous RL agent for resolving production infrastructure incidents 3.1× faster than human L1.

  • Gymnasium environment simulating 9 incident archetypes with confidence-gated observations.
  • Achieved 73% P1 resolution rate across 500+ episodes without relying on LLM judges.
PyTorch Gymnasium PPO/DQN Llama-3.3-70B
CASE_STUDY: OPTI-FAB

OPTI-FAB

Semiconductor Edge AI

Stream-aware edge inference pipeline for real-time 300mm wafer defect inspection.

  • TensorRT FP16 quantised MobileNetV2 achieving 1,262 FPS at 0.79ms latency.
  • 63.9% latency reduction with <1% accuracy drop across 8 defect classes.
TensorRT MobileNetV2 CUDA
CASE_STUDY: CROP_YIELD

Crop Yield

XGBoost + SHAP Explainability

End-to-end predictive pipeline on 18,000+ climate and soil records for agricultural forecasting.

  • Optimised XGBoost achieving R²=0.98 across 6 crop varieties.
  • SHAP-powered dashboard for real-time, explainable yield predictions.
XGBoost SHAP Streamlit

Achievements

🏆

8+ Hackathons & Meta Finalist

Meta PyTorch OpenEnv 2026 Finalist (Easy 0.906 · Medium 0.887 · Hard 0.650). Competed in Microsoft Imagine Cup and UIDAI Data Hackathon, producing 5 production-style pipelines under intense deadlines.

👥

Team Leadership & Mentorship

Led 6+ hackathon teams as Technical Lead. Mentored a national hackathon team to selection by architecting the solution, preparing technical documentation, and training teammates through end-to-end design.

⚙️

7+ Production Pipelines

Built and deployed 7+ real-world intelligent systems, including Reinforcement Learning environments, stream-aware edge AI engines, and enterprise medallion data lakehouses handling 200K+ workloads.

STACK

Technical Skills

95%
Python
95%
JavaScript
80%
SQL
95%
HTML
90%
CSS
70%
C
70%
C++

AI & Machine Learning

PyTorchGymnasiumPPODQNTensorRT FP16CUDAGemma 4ReAct AgentsUnsloth LoRAHugging FaceRLLangGraph

Pipelines & Data Engineering

PySpark 3.5SparkApache AirflowKafkaDatabricksDelta Lake (B→S→G)ETL/ELTData LineageSchema EnforcementParquetData Quality

Backend & Cloud Infra

FastAPIDockerDocker ComposePostgreSQL 15AWS (S3/IAM)GitHub Actions CI/CDSQLAlchemy ORMRESTful APIsPydanticpytestLinux

Data Science & Analytics

PythonPandasNumPyScikit-learnSHAPPower BIXGBoostStatistical ModellingStreamlit
WRITING

Blog

Architecture

How we built confidence-gating into IncidentMind

IncidentMind deep dive · Reinforcement Learning · System Design
Systems

Why 70% of inference latency has nothing to do with the model

OPTI-FAB insight · Edge AI · TensorRT
CAREER

Work Experience

Design Note #03

"Experience is presented as a high-density technical ledger. We avoid narrative-heavy descriptions in favor of Monospace metadata blocks and bulleted impact metrics, echoing the aesthetics of a system log."

REF_ID: LEDGER_PHILOSOPHY_01
Scion Research & Development
Data Science Analyst Intern
2025 · Jan – Aug (8 months)
[ STACK: Python, Pandas, SQLAlchemy, SQL, Power BI, Scikit-learn ]
FOUNDATION

Education

Bon Secours College for Women
B.Sc. Data Science (CGPA: 9.3/10)
2023 – 2026 · Bharathidasan University · NAAC A++
Certifications
Google Data Analytics & Salesforce AgentBlazer Champion
Professional Credentials
NxtWave
Full Stack Development Program
2024 – Present · Practical Web & API Architecting
Al-Mubeen Matriculation School
Higher Secondary (12th Standard)
Score: 80.17% · Foundation Studies
CONNECT

Contact

zsh — 120x40
asmitha-m --status
[ STATUS: AVAILABLE_FOR_HIRE ]
[ EMAIL: asmitha8825@gmail.com ]
[ INTERESTS: Data Engineer, Full Stack Developer, AI Engineer, Data Scientist ]
asmitha-m --contact
GitHub LinkedIn Email
Currently open to: AI engineering internships · Data engineering roles · Full-stack development · Research roles