Sri Nikitha Veerla

IT User Support Services — Texas A&M University–Corpus Christi

• Automated log parsing & diagnostics with Python and C++; reusable CLI tools.
• SQL jobs and dashboards to improve Argos / DegreeWorks data quality and visibility.
• Runbooks and knowledge base articles that reduced escalations and onboarding time.
• Ticket workflow insights that highlighted SLA risks and reduced MTTR.
• Standardized monitoring snippets to surface common failure modes quickly.

AI/ML Backend Development Intern — Tara Infotech

• Built and optimized backend APIs integrating ML inference endpoints.
• Containerized model services with safe rollouts; added telemetry for RCA.
• Validation and rate-limiting patterns to protect high-traffic routes.
• Reduced p95 latency with schema/index tuning and caching strategies.
• Improved CI stability with tests and static checks for critical paths.

Web Development Intern — Oasis Infobyte

• Developed responsive UI components; predictable state and secure forms.
• Query/index tuning and code-splitting to improve perceived performance.
• REST endpoints with clear error contracts and input sanitization.
• Session/token auth hardening and role-based access patterns.
• Tests & lint rules in CI to keep main stable during rapid iteration.

Multi-Layered AI Framework for Spoofing Detection

Real-time spoof detection using a hybrid ensemble of classic ML and transformer encoders. Designed for production with containerized services, health checks, and human-in-the-loop review.

• Ensemble architecture (SVM, Isolation Forest, transformer encoder).
• Event-driven inference, health probes, and drift alerting.
• Review queue and threshold tuning for operations.

• ~95% detection accuracy during evaluation.
• >60% throughput gain after optimization.
• p95 latency under target budget with caching.

Scalable Distributed System with AWS

Message-driven system for bursty workloads with schema validation, failure isolation, and autoscaling tuned to processing lag. Built with operational dashboards for fast incident response.

• Schema validation, DLQs, and replay tools for recovery.
• Autoscaling based on consumer lag and queue depth.
• Operational dashboards with alerts for SLOs.

• 500K+ requests handled in test runs.
• ~40% improvement in p95 latency vs baseline.
• High anomaly detection rate during pilots.

AI-Powered Weather Prediction System

Forecasting platform with an end-to-end pipeline: feature store, training, registry, and a public API backed by rate limiting and caching. Visual dashboard for analysts and developers.

• Feature store and versioned model registry.
• REST API with caching and quota controls.
• Analyst dashboard with exportable reports.

• ~98% forecast accuracy on holdout data.
• 1M+ monthly API calls during peak periods.
• ~25% faster responses after caching rollout.

Wastewater Flow Anomaly Prediction

Multivariate anomaly detection on wastewater flow signals to surface operational outliers for early intervention.

• Seasonal and diurnal feature engineering.
• LSTM baseline with attention variants.
• Graph-based edges for sensor proximity.

• Higher recall on rare anomalies vs statistical baselines.
• Reduced false positives via threshold calibration.
• Stable inference across seasonal changes.

Energy Optimization Experiments

Forecasting and optimization notebooks that balance demand, cost, and efficiency for scenario planning.

• ARIMA/Prophet baselines with exogenous features.
• Feature pipelines for weather and usage data.
• Scenario evaluation for peak shaving strategies.

• Lower MAPE vs naive splits in experiments.
• Demonstrated peak reduction potential.
• Reproducible runs with seeds and reports.

Parallel kNN (MPI)

Distributed k-Nearest Neighbors to scale distance computations across nodes for large datasets.

• Data partitioning with balanced work distribution.
• Collective operations to gather nearest neighbors.
• Benchmark harness for reproducible tests.

• Wall-clock speedups with node count increases.
• Near-linear scaling at moderate sizes.
• Reduced memory pressure via chunking.

Dark Tracer: Early Detection Framework for Malware (IJECS, 2023)

An early-warning malware pipeline using syscall and process-tree signals with a focus on explainability and low resource consumption for edge scenarios.

• Signal engineering across syscalls, process trees, and heuristics.
• Model comparisons: SVM, Random Forest, and boosted methods.
• Runtime profiling to balance precision with memory footprint.

• High precision/recall on curated malware sets.
• Low memory usage suitable for constrained devices.
• Stable detection under mild obfuscation.

Emotion-Based Music Player (IJARS, 2023)

Affect-aware recommendation approach that maps detected mood to track attributes, targeting a compact model suitable for near real-time inference.

• Augmented datasets to stabilize training for small models.
• Mood → tempo/key/energy mapping strategy for curation.
• User feedback loop to adapt preferences over time.

• Above-baseline classification accuracy on test sets.
• Reduced perceived mismatch in A/B user studies.
• Low-latency on-device inference in demos.

Second Place — NASA Space Apps Hackathon

Built an MVP leveraging NASA datasets/APIs to visualize earth observation data and flag weather anomalies with confidence scores.

• Interactive mapping dashboard with filters and drill-downs for analysts.
• Streamlined ingestion pipeline with data validation and caching.
• Clear roadmap for scaling compute/storage and improving UX.
• Collaborated across roles to deliver demo-ready features in 24 hours.
• Positive judge feedback on clarity, impact, and feasibility.

Entrepreneurship Development Cell (EDC) — JNTU Hyderabad

Promoted innovation and entrepreneurship through student events and mentorship—supporting idea sprints, MVPs, and hackathons.

• Organized startup idea competitions, hackathons, and awareness sessions.
• Collaborated with mentors and industry advisors for guidance.
• Built a supportive environment for startups and innovation activities.

AWS Certified (Honeywell Program)

• Containerized deployments on ECS/Fargate with blue/green rollouts and health checks.
• Least-privilege IAM roles, Secrets Manager/SSM Parameter Store for configuration.
• CloudWatch dashboards/alarms and X-Ray traces for latency/error budget monitoring.
• API Gateway + WAF in front of public services with throttling and request validation.
• Event-driven ETL using S3 + EventBridge + Lambda for asynchronous workloads.