Software Engineer · ML Researcher
Dhairya
Mishra
Building and deploying AI/ML production systems at scale — multimodal pipelines, LLM tooling, and distributed data infrastructure. 5+ years shipping computer vision, NLP, and cloud-native services.
Affiliated With
Featured Projects
Research and production work spanning ML, full-stack, and cloud systems.
Multiplayer Frames
Solaris — Multiplayer Video World Model in Minecraft
First multiplayer video world model generating consistent first-person observations for two players simultaneously, trained on 12.6M frames of coordinated Minecraft gameplay. Published on arXiv, NYU.
Shipped
Teserax.io — Graph-Based AI Thinking Tool
A dual-lane, chat-first exploration tool that transforms linear LLM chat into a visual, non-linear graph canvas with AI orchestration, crosslink reasoning, multi-model BYOK support, and cloud persistence — shipped v2.0 with 64 issues closed across 5 development phases.
Accuracy
Cloud NLP Classification on GCP
Production-ready multi-model text classification service with zero-downtime model switching, deployed on GCP Compute Engine. DistilBERT trained to 96.57% accuracy on a 24,783-sample dataset; 326+ test suite at 100% pass rate.
Accuracy
MRI Brain Tumor Detection & Segmentation
Multimodal MRI classification and segmentation model trained on BraTS dataset with shared encoder, achieving 91.3% accuracy and 97.1% sensitivity.
Skills & Technologies
Full-stack proficiency across ML/AI, cloud infrastructure, and modern web frameworks.
Languages
ML / AI
Cloud & DevOps
Frameworks & App Dev
Data & Storage
Testing & Observability
Experience Timeline
AI Engineer
Jan 2026 — PresentEvidenza · Brooklyn, NY
- Automated ingestion of legacy Human-vs-AI survey spreadsheets into a schema-validated database via a 6-stage Spark ETL, producing 7,500+ JSONL records across 26 domains and powering a new enterprise customer segmentation feature.
- Productionized AlphaEvolve persona generation system to synthesize diverse survey respondents at scale (25 new personas/week, 2 runs/week), with trait-coverage guardrails achieving >80% Monte Carlo coverage.
- Built competitive ad intelligence pipeline cataloging 1,200+ customer ads across 5 B2B verticals with semantic/sentiment + multimodal scoring; contributed to a 34% lift in engagement on curated ads/creatives.
- Engineered a multimodal creative feature-extraction pipeline using SAM-style segmentation, object detection, and video/image processing to compute 12+ per-ad signals (dominant color, objects, people count, duration); achieved 95% feature coverage.
- Built Spark-based ETL for recurring backfills and heavy joins across YouTube metadata and extracted creative signals; centralized 500 B2B ads in MySQL (Hive) and trained performance models for CTR, engagement, and completion rate.
- Defined human-vs-synthetic evaluation gates using a 50-user blind audit set; enforced ≥65% agreement as the production acceptance threshold for all persona outputs.
Sr. Software Development Engineer
Jan 2023 — Jan 2025CVS Health · New York, NY
- Shipped and reviewed 150+ PRs and owned on-call for CPE Webcore (Digital-Blocks 2.0, Previews, Experience Builder) serving millions of customers daily; achieved SLA compliance and reduced downtime by 12-15% on customer-facing core platform systems.
- Led CI/CD migration from GitLab to GitHub EMU under CPE web-core standards; migrated all accessibility applications and pipelines, validated dependencies, decommissioned legacy Jenkins/ArgoCD deployments, and trained teammates — achieved 100% adoption and reduced technical debt.
- Enhanced observability by provisioning GCS storage, Postgres databases, and a custom OpenTelemetry-to-Grafana pipeline simulating diverse request scenarios; improved median incident MTTR by 25-30% and safeguarded service stability during scale events.
- Piloted AI-powered Image-to-Alt-Text solution using Hugging Face transformers, establishing production-grade pipelines for enterprise rollout; reduced downstream defects by 15%. Origin: 1st Prize, CVS GenAI Hackathon (Aug 2023).
- Delivered RAG chatbot prototype leveraging OpenAI + ChromaDB with secured governance and Slack integration for self-serve troubleshooting; reduced manual ticket resolution time by 20%. Hosted on CVS Enterprise GCP clusters via Terraform and ArgoCD.
- Automated accessibility QA by integrating axe-core + Playwright into GitHub Actions pipelines; shift-left automation cut production accessibility issues by 60-90% (varying by application maturity) and streamlined WCAG violation detection across teams.
- Established the microservice ecosystem and GCP cloud infrastructure for the accessibility organization — the foundation for all team tooling, reporting, and compliance workflows.
- Launched A11yScore PoC with relational queries and leadership dashboards surfacing accessibility metrics; shortened defect resolution time by 20% and guided engineering resource re-allocation that increased throughput by 25%.
- Provisioned MongoDB database and APIs for the Accessibility org (replacing Rally as quasi-database) with optimized schema, auth integration, and front-end interfaces; boosted data transaction speeds by 55%.
- Refactored Testaro and Next Reporter, combining 1,000+ accessibility test outputs from multiple tools into standardized dashboards; reduced report analysis time by 20% and expanded adoption across all accessibility teams.
- Designed weekly multi-project automation auditing every work-item tree in the CVS kanban-hybrid (Rally via Broadcom); improved delivery velocity by 90% and boosted edge case remediation.
Advanced Software Developer
Feb 2022 — Jan 2023Aetna Health · New York, NY
- Developed accessibility compliance automation suite (CAT, RallyScore, ThemeScore) covering test-case generation, Rally hierarchy scoring, and theme-level validation; reduced QA time by 75%.
- Designed Rally kanban migration pipeline for 600+ nested structures with schema validation and statistical checks; reduced migration time by 80-90% (from 4 weeks to 16 business hours).
- Provisioned encrypted API microservices for MongoDB access; boosted data transaction speeds by 55%.
- Built auto-authentication and platform health-check microservices, improving uptime and reducing manual credential management for internal apps.
- Implemented standardized Rally rulesets and automated validation checks; improved automation efficiency by 30% and reduced manual organizational errors by 25%.
- Modularized test templates to accelerate QA cycles; served as Viability Reviewer evaluating and approving new accessibility tools for the org.
Education
New York University
Courant Institute
M.S. Computer Science (AI)
Trine University
B.S. Software Engineering & Mathematics
Let's Build Something
Interested in collaborating on ML research, full-stack systems, or production AI? I'd love to hear from you.