AVAILABLE FOR AI/ML & FULL-STACK|PUNE, IN
> whoami --aliasAvadhoot

SHREYASH DAWARE

AI/ML ENGINEER & SYSTEMS BUILDER

Building AI systems that see, reason, and act. Deployed end-to-end wildlife tracking across 19 classes with YOLOv8 + ByteTrack at 48 FPS (cutting duplicates by 70%), built multi-agent swarm studios, and engineered production full-stack systems.

CV↓
⚲PUNE, MAHARASHTRA, INDIA
GitHub↗LinkedIn↗Twitter↗
Shreyash Daware (Avadhoot) - AI/ML Engineer in Suit
AVADHOOT· LEAD
★ 9.42 CGPA
SHREYASH DAWAREAI/ML
ONLINE
*SYSTEMS THAT SEE & ACT
9.42/ 10 CGPA
B.E. Artificial Intelligence & Data Science
GATEQualified
National Engineering Entrance · 2026
OpenSource
Contributor · Sematica
End-to-EndBuilder
Vision, RAG & Agents to Full-Stack Interfaces
SCROLL TO EXPLORE↓
02 // OPERATOR FILE·BACKGROUND & PHILOSOPHY

FROM DATA INSIGHTS.
TO PRODUCTION SYSTEMS.

Bridging machine learning research with robust engineering: low-latency vision inference, multi-agent frameworks, and dependable full-stack applications.

// ORIGIN & FOCUS

I'm a final-year Artificial Intelligence & Data Science student at M.M. College of Engineering, Pune. During my internship at xLogia Tech, I designed and deployed an end-to-end real-time wildlife detection & counting pipeline across 19 animal classes with YOLOv8 and ByteTrack multi-object tracking on 10,000+ camera images.

By implementing persistent tracking IDs across video frames, our pipeline eliminated double-counting, reducing duplicate counts by 70% while streaming at 48 FPS.

My philosophy centres on full-cycle ownership: writing clean, vectorized model code in PyTorch, exposing high-throughput asynchronous APIs with FastAPI, containerizing services with Docker, and building intuitive, accessible frontends with Next.js and TypeScript.

DOWNLOAD PDF (CV)↓
// ACADEMIC FOUNDATIONSPPU PUNE
DEGREE PROGRAM
B.E. Artificial Intelligence & Data Science
M.M. College of Engineering · Pune, Maharashtra
CUMULATIVE GPA
9.42 / 10.0
TOP TIER
Expected Aug 2027
NOTABLE RECOGNITION
  • ▸First place, Poster Presentation Competition (Pi Day) — Shri Savitribai Phule University.
  • ▸Consolation prize, Internal Smart India Hackathon 2025 (college level).
  • ▸GATE 2026 Qualified in Artificial Intelligence & Data Science.
// LEADERSHIP & COMMUNITYROLES
CoordinatorUdaan — student project showcase

Ran the show end to end: logistics, external evaluators, winner selection.

TreasurerNEURA Technical Club

Budgeting and allocation for a college-wide hackathon, plus sponsors and organisers.

// ENGINEERING METHODOLOGYSTANDARDS
01 / DATA INTEGRITY

Rigorous annotation & augmentation to conquer real-world distribution shift.

02 / LOW LATENCY

Tuned inference resolution and asynchronous streaming for 48 FPS throughput.

03 / END-TO-END

Complete lifecycle from model weights to containerized production APIs.

Computer VisionYOLOv8ByteTrackPyTorchFastAPINext.js 16TypeScriptDockerQdrant Vector DB
03 // TECHNICAL SPECIFICATIONS·CAPABILITIES & ARCHITECTURES

SYSTEMS ARCHITECTURE.

Battle-tested tools and frameworks across Computer Vision, Neural Architectures, Async Backends, and Cloud Infrastructure.

01 // COMPUTER VISION & PERCEPTIONPRODUCTION DEPLOYED

Object Detection & Multi-Object Tracking

End-to-end vision pipelines engineered for real-time video inference, edge deployment, and persistent object identity tracking.

YOLOv8 & YOLOv11Architecture & Weights
ByteTrack Multi-Object TrackingPersistent ID Matching
OpenCV Video PipelinesRTSP & Low Latency
Detection & SegmentationBounding Box Geometry
Synthetic AugmentationRoboflow & Albumentations
Low-Light Optimization+30% Low-Res Accuracy
VERIFIED BENCHMARK:48 FPS @ 1080p · 80% mAP50 on 19 Classes
02 // DEEP LEARNING & MULTI-AGENT SYSTEMSACTIVE RESEARCH

Neural Networks & Agent Orchestration

Neural model training, hyperparameter optimization, vector search indexing, and multi-agent collaborative workflows.

PyTorch Deep LearningCustom CNNs & Heads
Autonomous Agent SwarmsMultiplayer Telemetry
Qdrant Vector DatabaseDense Semantic Search
BM25 Hybrid FusionSparse + Dense Retrieval
Cross-Encoder Re-RankingContext Precision
NumPy & PandasVectorized Pipelines
VERIFIED BENCHMARK:<80ms P95 Latency · 10 Concurrent Agents
03 // FULL-STACK & ASYNCHRONOUS SYSTEMSPRODUCTION READY

Modern Web Architectures & APIs

High-throughput async APIs, WebSocket streaming servers, and responsive, accessible client-side applications.

Next.js 16 (App Router)Server & Client Comps
React 19 & TypeScriptStrict Type-Safety
FastAPI (Python)Async Event Loops
Tailwind CSS v4Modern Design Systems
WebSocket Real-Time SyncLow-Jitter Telemetry
RESTful API ArchitectureOpenAPI / Swagger
VERIFIED BENCHMARK:Sub-second Page Hydration · Zero-Lag Canvas
04 // INFRASTRUCTURE, PERSISTENCE & MLOPSSCALABLE FOUNDATION

Deployment, Caching & Cloud Tooling

Containerized deployment workflows, relational and vector databases, caching layers, and continuous integration.

Docker & MicroservicesContainer Orchestration
PostgreSQL & SQL/DBMSRelational Schemas
Redis In-Memory CachingHigh-Throughput State
Linux / POSIX ShellServer Administration
Git & GitHub ActionsAutomated CI/CD
Vercel PlatformEdge Compute Deployment
VERIFIED BENCHMARK:Isolated Containers · Sub-5ms In-Memory Reads
04 // SELECTED RUNS·PRODUCTION ARTIFACTS & SYSTEMS

FEATURED PROJECTS.

Production checkpoints, precision curves, inference metrics, and repository architectures.

3 OF 3 SYSTEMS ONLINE|VERIFIED REPOSITORIES
RUN // 01Wildlife-Detection-Pipeline
WILDLIFE DETECTION & COUNTING PIPELINE

WILDLIFE DETECTION & COUNTING PIPELINE

Real-Time Edge Computer Vision & Multi-Object Tracking

Engineered an end-to-end computer-vision pipeline for wildlife detection and counting across 19 animal classes on a 10,000+ camera-trap image dataset. Integrated ByteTrack to maintain persistent object IDs across frames, cutting duplicate counting by 70% while streaming at 48 FPS.

80%mAP@0.5
85%Precision
48 FPSThroughput
19 SpeciesClasses
YOLOv8ByteTrackPyTorchOpenCVFastAPIPython
VIEW CODE↗
RUN // 02agent-api
AGENT-API // COLLABORATIVE AI STUDIO

AGENT-API // COLLABORATIVE AI STUDIO

Multi-Agent Telemetry & Real-Time Multiplayer Canvas

Autonomous multi-agent orchestration engine where 10 specialized AI agents coordinate simultaneously to review, sketch, and assemble UI components with real-time multiplayer telemetry and zero-lag spatial canvas synchronization.

10 ConcurrentAgents
<15ms JitterTelemetry
Zero-LagCanvas
WebSocketsProtocol
Next.js 16TypeScriptWebSocketsAgentic AITailwind CSS
VIEW CODE↗
RUN // 03MedAssist
MEDASSIST // HYBRID ENTERPRISE RAG

MEDASSIST // HYBRID ENTERPRISE RAG

Clinical Vector Search & Cross-Encoder Re-Ranking

Production-grade retrieval-augmented generation engine pairing dense embeddings in Qdrant with BM25 sparse keyword search and cross-encoder re-ranking for ultra-precise clinical knowledge retrieval with sub-80ms P95 latency.

QdrantVector DB
BM25 HybridFusion
<80msLatency P95
Top-1 94%Accuracy
QdrantFastAPIPythonRAGCross-EncoderPyTorch
VIEW CODE↗
05 // CHRONOLOGY·CAREER PROGRESSION & RUNS

EXPERIENCE & HONORS.

A timeline of industry internships, commercial engineering, and academic achievements.

TIMELINE 2023 — 2027|CHRONOLOGICAL MILESTONES
2026Feb – Mar 2026
INDUSTRY INTERNSHIP

AI/ML Engineering Intern

xLogia Tech
  • ▸Engineered an end-to-end computer-vision pipeline for wildlife detection & counting across 19 animal classes on 10,000+ camera-trap images (80% mAP50, 85% precision).
  • ▸Implemented ByteTrack multi-object tracking to hold persistent IDs across frames, cutting duplicate counting by 70% for accurate animal population monitoring.
  • ▸Architected a real-time streaming analytics pipeline running at 40–50 FPS with bounding boxes and live species-wise count overlays.
  • ▸Tuned model confidence thresholds and inference resolutions so detection on low-quality, low-light footage improved by 30%.
PythonPyTorchYOLOv8ByteTrackOpenCVFastAPI
2025Jun 2025
COMMERCIAL CONTRACT

Corporate Web Developer

SKI Industrial
  • ▸Delivered a responsive corporate web platform for a bakery-machinery manufacturer across Home, About, Products, and Contact pages.
  • ▸Engineered fast loads, client-side routing, and SEO best practices on Vercel, achieving top-tier Core Web Vitals across mobile and desktop.
Next.jsTypeScriptTailwind CSSVercelSEO Optimization
20272023 – Expected Aug 2027
ACADEMIC HONORS & LEADERSHIP

B.E. in Artificial Intelligence & Data Science

M.M. College of Engineering
  • ▸Achieved a 9.42 CGPA in Bachelor of Engineering coursework in Artificial Intelligence & Data Science at Savitribai Phule Pune University.
  • ▸Won 1st Place in Poster Presentation Competition (Pi Day) at Savitribai Phule Pune University.
  • ▸Qualified GATE 2026 examination in Artificial Intelligence & Data Science.
  • ▸Served as Treasurer of NEURA Technical Club (budgeting and sponsorships for hackathons) and Coordinator for the Udaan annual student project showcase.
9.42 / 10 CGPAGATE 2026 QualifiedPi Day 1st PlaceNEURA Club Treasurer
06 // DISPATCH·TRANSMISSION & COLLABORATION

LET'S WORK
TOGETHER.

Open for full-time AI/ML engineering roles, computer vision pipelines, and full-stack software development.

Have a computer vision challenge, an autonomous agent pipeline, or a full-stack platform to engineer? I'm currently open to full-time engineering roles, high-impact freelance projects, and research collaborations.

START A CONVERSATION↗
EMAIL:daware.shreyash@gmail.com
GitHub:noNameDS09↗
LinkedIn:shreyashdaware↗
Twitter:noNameDS09↗
// LIVE OPERATOR TELEMETRYONLINE
LOCAL TIME · PUNE (IST)
16:00:00 PM
UTC+05:30 · ASIA/KOLKATA
BASE COORDINATES:Pune, Maharashtra, India
SPECIALIZATION:AI/ML & FULL-STACK
RESPONSE WINDOW:< 24 HOURS
CURRENT STATUS:READY FOR RUNS
© 2026 Shreyash Daware (Avadhoot). All rights reserved.
Next.js 16 · React 19 · Tailwind CSS · Swarm Cursors