-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
330 lines (296 loc) · 21.7 KB
/
Copy pathindex.html
File metadata and controls
330 lines (296 loc) · 21.7 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Abhishek Chintagunta — AI Systems Engineer</title>
<meta name="description" content="Abhishek Chintagunta — AI systems architect & CTO. Agentic AI, LLM orchestration, multi-tenant RAG, voice AI. PhD, Aerospace Engineering.">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link href="https://fonts.googleapis.com/css2?family=Space+Grotesk:wght@500;700&family=Inter:wght@400;500;600;700&family=JetBrains+Mono:wght@400;500&display=swap" rel="stylesheet">
<link rel="stylesheet" href="assets/style.css">
<link rel="stylesheet" href="assets/scene3d.css">
<script src="assets/theme.js"></script>
</head>
<body>
<header class="hero">
<div id="hero3d" aria-hidden="true"></div>
<nav class="nav">
<span class="nav-logo">chintagunta<span class="dot">/</span>abhishek</span>
<div class="nav-links">
<a href="#work">Work</a>
<a href="#experience">Experience</a>
<a href="#projects">Projects</a>
<a href="#articles">Articles</a>
<a href="#research">Numerical Computing</a>
<a href="#stats">Stats</a>
<a href="#stack">Stack</a>
<a href="#activity">Activity</a>
<a href="#contact">Contact</a>
<button id="theme-toggle" class="theme-btn" aria-label="Toggle dark/light theme">◑</button>
</div>
</nav>
<div class="hero-content">
<img class="avatar" src="https://github.com/chintagunta.png?size=200" alt="Abhishek Chintagunta" onerror="this.style.display='none'">
<p class="eyebrow">⚡ AI Systems Architect · ex-CTO/Chief AI Officer, Slikk AI · Hyderabad, India</p>
<h1>I build things<br>that <span class="flip"><span class="flip-inner">decide.</span></span></h1>
<p class="tagline">Agentic AI platforms. LLM orchestration. Enterprise systems that act on their own.
PhD in computational fluid dynamics, because shockwaves came first.</p>
<div class="hero-actions">
<a class="btn primary" href="https://github.com/chintagunta" target="_blank" rel="noopener">GitHub →</a>
<a class="btn" href="https://www.linkedin.com/in/abhishekchintagunta/" target="_blank" rel="noopener">LinkedIn</a>
<a class="btn" href="#contact">Contact</a>
</div>
</div>
<div class="marquee">
<div class="marquee-track">
<span>AGENTIC AI</span><span>LLM ORCHESTRATION</span><span>MULTI-AGENT SYSTEMS</span><span>PYTHON</span>
<span>CUDA</span><span>MACHINE LEARNING</span><span>CFD</span><span>NESTJS</span><span>SQL</span>
<span>AGENTIC AI</span><span>LLM ORCHESTRATION</span><span>MULTI-AGENT SYSTEMS</span><span>PYTHON</span>
<span>CUDA</span><span>MACHINE LEARNING</span><span>CFD</span><span>NESTJS</span><span>SQL</span>
</div>
</div>
</header>
<main>
<section id="work" class="section">
<h2><span class="kicker">01</span>Problems I've cracked</h2>
<div class="case-list">
<article class="case">
<div class="case-head">
<h3>Agentic AI platform: multi-agent orchestration, voice, and multi-tenant RAG</h3>
<span class="case-tag">Slikk AI · CTO / Chief AI Officer · 2025–2026</span>
</div>
<p class="case-problem"><strong>Problem:</strong> Enterprises need software that doesn't just answer
questions but takes actions — drafts legal documents, authors leases, runs reconciliations — across text
and voice, for many tenants, without leaking one customer's data into another's context.</p>
<p class="case-solution"><strong>Approach:</strong> Architected a multi-agent orchestration framework of
autonomous "employee agents" (including specialized legal agents for document drafting and lease
authoring); a first-of-its-kind single-pane collaborative interface unifying human-to-human and
human-to-AI chat with built-in voice assistants, configurable RAG pipelines, and an analytics engine; and
a secure multi-tenant RAG architecture built from the ground up with strict per-tenant data isolation
serving both text and voice. Ran it all on hybrid-cloud microservices across AWS and Azure (NestJS,
Python, ReactJS, React Native).</p>
<p class="case-result"><strong>Result:</strong> Led delivery end-to-end — including a voice-first platform
connecting and empowering the blue-collar workforce, where voice interfaces drove accessibility — turning
complex business use cases into production architecture that shipped.</p>
</article>
<article class="case">
<div class="case-head">
<h3>LLM-driven case processing & report authoring for Oracle Argus</h3>
<span class="case-tag">Ultragenic R&T · AI/ML Architect · 2022–2025</span>
</div>
<p class="case-problem"><strong>Problem:</strong> Pharmacovigilance case processing and periodic safety
report authoring are high-volume, high-precision tasks traditionally done by hand inside Oracle Argus —
slow, expensive, and error-prone at scale.</p>
<p class="case-solution"><strong>Approach:</strong> Led AI/ML architecture across three systems — UltraIntake
(automated case ingestion and triage), UltraARM (LLM-assisted periodic report authoring), and UltraMig
(migration tooling for legacy Argus data). Designed the ML pipelines and led the team building them.</p>
<p class="case-result"><strong>Result:</strong> Cut manual processing time on case intake and report drafting
while keeping outputs auditable enough for regulated pharma workflows.</p>
</article>
<article class="case">
<div class="case-head">
<h3>Low-latency algorithmic trading platform</h3>
<span class="case-tag">Sprouts Pvt. Ltd · Co-Founder & CTO · 2015–2017</span>
</div>
<p class="case-problem"><strong>Problem:</strong> Retail-accessible algo trading needed backtesting, paper
trading, and live execution on the same codebase, plus a news feed fast enough to act as a trading signal
rather than a headline ticker.</p>
<p class="case-solution"><strong>Approach:</strong> Built the platform's backtesting, paper, and live trading
modes in Python, and a separate low-latency news aggregator (Python, Tkinter, SQL, BeautifulSoup, LXML,
multiprocessing) that scraped and ranked signal-relevant news faster than standard feed delays allowed.</p>
<p class="case-result"><strong>Result:</strong> A working trading stack covering the full strategy lifecycle
— from backtest to live order execution.</p>
</article>
<article class="case">
<div class="case-head">
<h3>Dispersion-improved numerical scheme for computational aeroacoustics</h3>
<span class="case-tag">PhD, Queen Mary University of London · 2014–2019</span>
</div>
<p class="case-problem"><strong>Problem:</strong> The CABARET finite-difference scheme, used for simulating
wave propagation and shock-boundary interactions, suffered from dispersion error that degraded accuracy on
oblique and multi-dimensional wave problems.</p>
<p class="case-solution"><strong>Approach:</strong> Developed a flux-corrected, dispersion-improved
modification to the scheme (MRCAB-DISP) and validated it against standard CABARET on 2D oblique wave cases
and jet noise problems, building on GPU-parallel unstructured CFD work from an earlier MPhil at Surrey.</p>
<p class="case-result"><strong>Result:</strong> Published in <em>Journal of Computers & Fluids</em> and
presented at two AIAA/CEAS Aeroacoustics Conferences.</p>
</article>
</div>
</section>
<section id="experience" class="section">
<h2><span class="kicker">02</span>Experience</h2>
<div class="timeline">
<div class="tl-item">
<span class="tl-when">2025 – 2026</span>
<div class="tl-body"><h3>CTO / Chief AI Officer · Slikk AI</h3>
<p>Owned AI and platform architecture end-to-end: multi-agent orchestration framework of autonomous employee agents (incl. legal drafting & lease authoring), unified voice + text collaborative interface with configurable RAG pipelines, secure multi-tenant RAG with per-tenant isolation, and hybrid-cloud microservices on AWS + Azure. Led delivery of a voice-first platform for the blue-collar workforce.</p></div>
</div>
<div class="tl-item">
<span class="tl-when">2022 – 2025</span>
<div class="tl-body"><h3>Associate Principal / Senior Consultant · Ultragenic R&T</h3>
<p>AI/ML architect and team lead for Oracle Argus pharmacovigilance products: UltraIntake (case processing), UltraARM (LLM report authoring), UltraMig (migration).</p></div>
</div>
<div class="tl-item">
<span class="tl-when">2019 – 2022</span>
<div class="tl-body"><h3>Assistant Professor · Vellore Institute of Technology</h3>
<p>Taught and researched numerical methods, CFD, and applied ML.</p></div>
</div>
<div class="tl-item">
<span class="tl-when">2018 – 2019</span>
<div class="tl-body"><h3>Co-Founder & CTO · Dieatto Wellness</h3>
<p>Built an AI-based nutrition recommendation engine.</p></div>
</div>
<div class="tl-item">
<span class="tl-when">2015 – 2017</span>
<div class="tl-body"><h3>Co-Founder & CTO · Sprouts Pvt. Ltd</h3>
<p>Chief architect for a low-latency algorithmic trading platform — backtesting, paper, and live execution in Python.</p></div>
</div>
<div class="tl-item">
<span class="tl-when">2014 – 2019</span>
<div class="tl-body"><h3>PhD, Aerospace Engineering · Queen Mary University of London</h3>
<p>Dispersion-improved CABARET schemes for computational aeroacoustics. Published in <em>Journal of Computers & Fluids</em>; two AIAA/CEAS conference papers.</p></div>
</div>
<div class="tl-item">
<span class="tl-when">2010 – 2014</span>
<div class="tl-body"><h3>Postgraduate Researcher (MPhil) · University of Surrey</h3>
<p>GPU-parallel 3D unstructured CFD algorithms for turbomachinery, in collaboration with Rolls-Royce.</p></div>
</div>
</div>
</section>
<section id="projects" class="section">
<h2><span class="kicker">03</span>Open source</h2>
<p class="muted">Self-initiated tools around agentic AI safety, observability, and the plumbing that makes LLM systems work in production.</p>
<article class="proj">
<h3><a href="https://github.com/chintagunta/qfire-hipaa-firewall" target="_blank" rel="noopener">qfire-hipaa-firewall</a></h3>
<span class="proj-tag">Python · MIT · prompt security</span>
<p>A declarative, positive-security <strong>prompt firewall</strong> for LLM applications, with a HIPAA/PHI rule set. Rules and chains are authored as YAML; the engine evaluates them cheapest-first with short-circuiting, fails closed on any detector error, and writes an auditable decision trace for every evaluation.</p>
<p>Includes an optional local ONNX classifier node, so PHI-exfiltration attempts (patient names, MRNs, diagnoses headed to personal email) get blocked before they ever reach the model.</p>
</article>
<article class="proj">
<h3><a href="https://github.com/chintagunta/nosql-trace" target="_blank" rel="noopener">nosql-trace</a></h3>
<span class="proj-tag">Python · PyPI · agent observability</span>
<p>A durable, pluggable <strong>NoSQL log sink for agentic workflows</strong>. Records LLM calls, tool calls, and agent steps into MongoDB (or DocumentDB / Cosmos DB / FerretDB) or DynamoDB — logs survive process crashes, never block the agent on network I/O, and never raise into application code.</p>
<p><code>log_event()</code> does one local SQLite insert on the calling thread; buffering, retry/backoff, ULID generation, and serialization are all stdlib. Only two runtime dependencies: <code>pymongo</code> and <code>boto3</code>.</p>
</article>
<article class="proj">
<h3><a href="https://github.com/chintagunta/docx-to-prosemirror" target="_blank" rel="noopener">docx-to-prosemirror</a></h3>
<span class="proj-tag">Python · zero dependencies · document conversion</span>
<p>Converts Word OpenXML (<code>.docx</code>) documents to <strong>ProseMirror / Tiptap JSON</strong> with high fidelity. Most converters round-trip through Markdown or HTML and lose structure; this one parses <code>document.xml</code>, <code>styles.xml</code>, and <code>numbering.xml</code> directly and rebuilds ProseMirror's nested tree from Word's flat run model.</p>
<p>Stdlib only — <code>xml.etree.ElementTree</code> and <code>zipfile</code>. No required dependencies at all.</p>
</article>
<article class="proj">
<h3><a href="https://github.com/chintagunta/dynamic-langraph" target="_blank" rel="noopener">dynamic-langraph</a></h3>
<span class="proj-tag">Python · agent orchestration</span>
<p>Define <strong>LangGraph agent workflows in JSON</strong> instead of hardcoded graph topologies. DLG handles validation, compilation, resilience, and execution — change an agent's workflow by editing configuration, not redeploying code.</p>
</article>
<article class="proj">
<h3><a href="https://github.com/chintagunta/sql-query-tagger" target="_blank" rel="noopener">sql-query-tagger</a></h3>
<span class="proj-tag">Python · PyPI · SQL security</span>
<p>Static <strong>SQL query classifier and injection-risk analyzer</strong> for all RDS engines (PostgreSQL, MySQL/MariaDB, Aurora, Oracle, SQL Server). Given a query string, returns its type (DDL/DML/DQL/…) and a LOW→CRITICAL risk level with the specific patterns that triggered it — injection patterns, stacked queries, destructive DDL, engine-specific dangerous functions.</p>
<p>Purely static analysis over the query text — no database connection required. Useful as a guardrail in front of LLM-generated SQL.</p>
</article>
</section>
<section id="articles" class="section">
<h2><span class="kicker">04</span>Articles & whitepapers</h2>
<p class="muted">Research and engineering write-ups from my tenure leading AI at Slikk. Full papers at <a href="https://slikk.ai/whitepapers" target="_blank" rel="noopener">slikk.ai/whitepapers →</a></p>
<article class="proj">
<h3><a href="https://www.slikk.ai/whitepapers/rag-architecture" target="_blank" rel="noopener">High-Performance RAG Architecture: Production-Grade Knowledge Systems for the Enterprise</a></h3>
<span class="proj-tag">Enterprise AI · Jan 2026 · 15 min read</span>
<p>A technical blueprint for enterprise-grade RAG: modular pipeline architecture, Markdown-first document normalization, semantic chunking, and hybrid retrieval over PGVector — grounded, secure AI responses with strict tenant isolation.</p>
</article>
<article class="proj">
<h3><a href="https://www.slikk.ai/whitepapers/hallucination-mitigation" target="_blank" rel="noopener">Secure Foundations: Mitigating Hallucinations in Enterprise AI</a></h3>
<span class="proj-tag">Enterprise AI · Feb 2026 · 12 min read</span>
<p>An evidence-grounded, multi-layered architecture for high-stakes reliability: knowledge grounding, agent-based verification, and behavioral alignment — how enterprise AI systems earn trust when a wrong answer has real cost.</p>
</article>
<article class="proj">
<h3><a href="https://www.slikk.ai/whitepapers/ai-summarization" target="_blank" rel="noopener">AI Summarization: Transforming Documents into Actionable Insights</a></h3>
<span class="proj-tag">AI & Machine Learning · Dec 2025 · 10 min read</span>
<p>How AI-driven summarization workflows extract key insights from large document sets — improving decision-making and cutting analysis time across finance, legal, and audit use cases.</p>
</article>
</section>
<section id="research" class="section">
<h2><span class="kicker">05</span>Numerical computing</h2>
<p class="muted">Research in high-order numerical schemes for computational aeroacoustics and compressible flows. 68 citations · h-index 3 · <a href="https://scholar.google.com/citations?user=ERVGgS0AAAAJ&hl=en" target="_blank" rel="noopener">Google Scholar →</a></p>
<article class="proj">
<h3><a href="https://scholar.google.com/citations?user=ERVGgS0AAAAJ&hl=en" target="_blank" rel="noopener">Flux-corrected dispersion-improved CABARET schemes for linear and nonlinear wave propagation problems</a></h3>
<span class="proj-tag">Computers & Fluids · 2018 · 47 citations · with S.E. Naghibi, S.A. Karabasov</span>
<p>The core PhD contribution: a flux-corrected modification of the CABARET scheme that improves dispersion properties on oblique and multi-dimensional wave problems, validated on linear and nonlinear propagation benchmarks.</p>
</article>
<article class="proj">
<h3><a href="https://scholar.google.com/citations?user=ERVGgS0AAAAJ&hl=en" target="_blank" rel="noopener">CABARET on rotating meshes</a></h3>
<span class="proj-tag">Applied Mathematics and Computation · 2023 · 14 citations · with I.A. Solntsev, A.P. Markesteijn, S.A. Karabasov</span>
<p>Extends the CABARET scheme to rotating mesh frameworks — enabling low-dissipation simulation of rotor and turbomachinery flows.</p>
</article>
<article class="proj">
<h3><a href="https://scholar.google.com/citations?user=ERVGgS0AAAAJ&hl=en" target="_blank" rel="noopener">A fourth-order CABARET scheme for Computational Aeroacoustics</a></h3>
<span class="proj-tag">2019 · with S.E. Naghibi, A.P. Markesteijn, S.A. Karabasov</span>
<p>Raises CABARET to fourth-order accuracy for aeroacoustic wave propagation, cutting dispersion and dissipation error at practical grid resolutions.</p>
</article>
<article class="proj">
<h3><a href="https://scholar.google.com/citations?user=ERVGgS0AAAAJ&hl=en" target="_blank" rel="noopener">Dispersion improved CABARET for computational aeroacoustics</a></h3>
<span class="proj-tag">23rd AIAA/CEAS Aeroacoustics Conference · 2017 · with A.P. Markesteijn, S.E. Naghibi, S.A. Karabasov</span>
<p>Conference presentation of the dispersion-improved scheme applied to jet noise and 2D oblique wave test cases.</p>
</article>
<article class="proj">
<h3><a href="https://scholar.google.com/citations?user=ERVGgS0AAAAJ&hl=en" target="_blank" rel="noopener">Implicit gradients based novel conservative numerical scheme for compressible flows</a></h3>
<span class="proj-tag">2021 · with A.S. Chamarthi, N. Hoffmann, H. Nishikawa, et al.</span>
<p>A conservative scheme using implicit gradient reconstruction for compressible flow simulation.</p>
</article>
</section>
<section id="stats" class="section stats-section">
<h2><span class="kicker">06</span>By the numbers</h2>
<div id="stats-grid" class="stats-grid">
<div class="stat"><span class="stat-num" data-target="0">0</span><span class="stat-label">Public repos</span></div>
<div class="stat"><span class="stat-num" data-target="0">0</span><span class="stat-label">GitHub followers</span></div>
<div class="stat"><span class="stat-num" data-target="0">0</span><span class="stat-label">Stars earned</span></div>
<div class="stat"><span class="stat-num" data-target="4">0</span><span class="stat-label">Published papers</span></div>
<div class="stat"><span class="stat-num" data-target="16">0</span><span class="stat-label">Years building software</span></div>
<div class="stat"><span class="stat-num" data-target="3">0</span><span class="stat-label">Companies founded / led AI</span></div>
</div>
</section>
<section id="stack" class="section">
<h2><span class="kicker">07</span>Stack</h2>
<div class="stack-grid">
<div class="stack-col">
<h4>AI / ML</h4>
<ul><li>LLM orchestration (LangGraph, multi-agent systems)</li><li>Agentic architectures</li><li>Machine Learning</li><li>OCR / document extraction</li></ul>
</div>
<div class="stack-col">
<h4>Languages</h4>
<ul><li>Python</li><li>JavaScript / TypeScript</li><li>C / C++</li><li>FORTRAN</li><li>SQL</li></ul>
</div>
<div class="stack-col">
<h4>Backend / Data</h4>
<ul><li>Node.js / Nest.js</li><li>Hadoop / Hive / Spark</li><li>CUDA / parallel computing</li><li>System integration & APIs</li></ul>
</div>
<div class="stack-col">
<h4>Other</h4>
<ul><li>Computational Fluid Dynamics</li><li>Numerical methods</li><li>Blockchain</li><li>EDA / Tableau</li></ul>
</div>
</div>
</section>
<section id="activity" class="section">
<h2><span class="kicker">08</span>Live activity</h2>
<p class="muted">Fetched client-side from the GitHub REST API on every page load.</p>
<div id="activity-feed" class="activity-feed">
<p class="loading">Fetching recent activity…</p>
</div>
</section>
<section id="contact" class="section contact-section">
<h2><span class="kicker">09</span>Let's talk</h2>
<p class="muted">Open to discussing agentic AI, enterprise automation, or numerical computing.</p>
<div class="hero-actions">
<a class="btn primary" href="https://github.com/chintagunta" target="_blank" rel="noopener">GitHub →</a>
<a class="btn" href="https://www.linkedin.com/in/abhishekchintagunta/" target="_blank" rel="noopener">LinkedIn</a>
</div>
</section>
</main>
<footer>
<p>© <span id="year"></span> Abhishek Chintagunta · Live data via GitHub REST API</p>
</footer>
<script src="assets/script.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/three.js/r128/three.min.js"></script>
<script src="assets/scene3d.js"></script>
</body>
</html>